From 66f8625a427415463201dee3d3c92c74fa76cdcd Mon Sep 17 00:00:00 2001 From: Kinar Date: Mon, 4 Dec 2023 04:10:34 -0800 Subject: [PATCH 1/9] Added Holistic Landmarker Python API --- mediapipe/python/BUILD | 1 + mediapipe/tasks/python/core/task_info.py | 8 +- mediapipe/tasks/python/test/vision/BUILD | 25 + .../test/vision/holistic_landmarker_test.py | 114 ++++ mediapipe/tasks/python/vision/BUILD | 23 + .../python/vision/holistic_landmarker.py | 567 ++++++++++++++++++ mediapipe/tasks/testdata/vision/BUILD | 4 + 7 files changed, 740 insertions(+), 2 deletions(-) create mode 100644 mediapipe/tasks/python/test/vision/holistic_landmarker_test.py create mode 100644 mediapipe/tasks/python/vision/holistic_landmarker.py diff --git a/mediapipe/python/BUILD b/mediapipe/python/BUILD index 4c89aa6c1c..e35422b855 100644 --- a/mediapipe/python/BUILD +++ b/mediapipe/python/BUILD @@ -103,6 +103,7 @@ cc_library( "//mediapipe/tasks/cc/vision/interactive_segmenter:interactive_segmenter_graph", "//mediapipe/tasks/cc/vision/object_detector:object_detector_graph", "//mediapipe/tasks/cc/vision/pose_landmarker:pose_landmarker_graph", + "//mediapipe/tasks/cc/vision/holistic_landmarker:holistic_landmarker_graph", ], ) diff --git a/mediapipe/tasks/python/core/task_info.py b/mediapipe/tasks/python/core/task_info.py index 1816d60e05..8941033613 100644 --- a/mediapipe/tasks/python/core/task_info.py +++ b/mediapipe/tasks/python/core/task_info.py @@ -82,8 +82,12 @@ def add_stream_name_prefix(tag_index_name): ) task_subgraph_options = calculator_options_pb2.CalculatorOptions() task_options_proto = self.task_options.to_pb2() - task_subgraph_options.Extensions[task_options_proto.ext].CopyFrom( - task_options_proto) + + # For protobuf 2 compat. + if hasattr(task_options_proto, 'ext'): + task_subgraph_options.Extensions[task_options_proto.ext].CopyFrom( + task_options_proto) + if not enable_flow_limiting: return calculator_pb2.CalculatorGraphConfig( node=[ diff --git a/mediapipe/tasks/python/test/vision/BUILD b/mediapipe/tasks/python/test/vision/BUILD index c6fae0e6cf..374ba689c1 100644 --- a/mediapipe/tasks/python/test/vision/BUILD +++ b/mediapipe/tasks/python/test/vision/BUILD @@ -194,6 +194,31 @@ py_test( ], ) +py_test( + name = "holistic_landmarker_test", + srcs = ["holistic_landmarker_test.py"], + data = [ + "//mediapipe/tasks/testdata/vision:test_images", + "//mediapipe/tasks/testdata/vision:test_models", + "//mediapipe/tasks/testdata/vision:test_protos", + ], + tags = ["not_run:arm"], + deps = [ + "//mediapipe/framework/formats:classification_py_pb2", + "//mediapipe/framework/formats:landmark_py_pb2", + "//mediapipe/python:_framework_bindings", + "//mediapipe/tasks/python/components/containers:category", + "//mediapipe/tasks/python/components/containers:landmark", + "//mediapipe/tasks/python/components/containers:rect", + "//mediapipe/tasks/python/core:base_options", + "//mediapipe/tasks/python/test:test_utils", + "//mediapipe/tasks/python/vision:holistic_landmarker", + "//mediapipe/tasks/python/vision/core:image_processing_options", + "//mediapipe/tasks/python/vision/core:vision_task_running_mode", + "@com_google_protobuf//:protobuf_python", + ], +) + py_test( name = "face_aligner_test", srcs = ["face_aligner_test.py"], diff --git a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py new file mode 100644 index 0000000000..0c91793013 --- /dev/null +++ b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py @@ -0,0 +1,114 @@ +# Copyright 2023 The MediaPipe Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +"""Tests for holistic landmarker.""" + +import enum +from unittest import mock + +from absl.testing import absltest +from absl.testing import parameterized +import numpy as np + +from google.protobuf import text_format +from mediapipe.framework.formats import classification_pb2 +from mediapipe.framework.formats import landmark_pb2 +from mediapipe.python._framework_bindings import image as image_module +from mediapipe.tasks.python.components.containers import category as category_module +from mediapipe.tasks.python.components.containers import landmark as landmark_module +from mediapipe.tasks.python.components.containers import rect as rect_module +from mediapipe.tasks.python.core import base_options as base_options_module +from mediapipe.tasks.python.test import test_utils +from mediapipe.tasks.python.vision import holistic_landmarker +from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module +from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module + + +HolisticLandmarkerResult = holistic_landmarker.HolisticLandmarkerResult +_BaseOptions = base_options_module.BaseOptions +_Category = category_module.Category +_Rect = rect_module.Rect +_Landmark = landmark_module.Landmark +_NormalizedLandmark = landmark_module.NormalizedLandmark +_Image = image_module.Image +_HolisticLandmarker = holistic_landmarker.HolisticLandmarker +_HolisticLandmarkerOptions = holistic_landmarker.HolisticLandmarkerOptions +_RUNNING_MODE = running_mode_module.VisionTaskRunningMode +_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions + +_HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE = 'face_landmarker.task' +_POSE_IMAGE = 'male_full_height_hands.jpg' +_CAT_IMAGE = 'cat.jpg' +_HOLISTIC_RESULT = "male_full_height_hands_result_cpu.pbtxt" +_LANDMARKS_MARGIN = 0.03 +_BLENDSHAPES_MARGIN = 0.13 + + +class ModelFileType(enum.Enum): + FILE_CONTENT = 1 + FILE_NAME = 2 + + +class HolisticLandmarkerTest(parameterized.TestCase): + + def setUp(self): + super().setUp() + self.test_image = _Image.create_from_file( + test_utils.get_test_data_path(_POSE_IMAGE) + ) + self.model_path = test_utils.get_test_data_path( + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE + ) + + @parameterized.parameters( + ( + ModelFileType.FILE_NAME, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE + ), + ( + ModelFileType.FILE_CONTENT, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE + ), + ) + def test_detect( + self, + model_file_type, + model_name + ): + # Creates holistic landmarker. + model_path = test_utils.get_test_data_path(model_name) + if model_file_type is ModelFileType.FILE_NAME: + base_options = _BaseOptions(model_asset_path=model_path) + elif model_file_type is ModelFileType.FILE_CONTENT: + with open(model_path, 'rb') as f: + model_content = f.read() + base_options = _BaseOptions(model_asset_buffer=model_content) + else: + # Should never happen + raise ValueError('model_file_type is invalid.') + + options = _HolisticLandmarkerOptions( + base_options=base_options + ) + landmarker = _HolisticLandmarker.create_from_options(options) + + # Performs holistic landmarks detection on the input. + detection_result = landmarker.detect(self.test_image) + + # Closes the holistic landmarker explicitly when the holistic landmarker is not used + # in a context. + landmarker.close() + + +if __name__ == '__main__': + absltest.main() diff --git a/mediapipe/tasks/python/vision/BUILD b/mediapipe/tasks/python/vision/BUILD index 0c1d42297d..8253a92327 100644 --- a/mediapipe/tasks/python/vision/BUILD +++ b/mediapipe/tasks/python/vision/BUILD @@ -243,6 +243,29 @@ py_library( ], ) +py_library( + name = "holistic_landmarker", + srcs = [ + "holistic_landmarker.py", + ], + deps = [ + "//mediapipe/framework/formats:classification_py_pb2", + "//mediapipe/framework/formats:landmark_py_pb2", + "//mediapipe/python:_framework_bindings", + "//mediapipe/python:packet_creator", + "//mediapipe/python:packet_getter", + "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_landmarker_graph_options_py_pb2", + "//mediapipe/tasks/python/components/containers:category", + "//mediapipe/tasks/python/components/containers:landmark", + "//mediapipe/tasks/python/core:base_options", + "//mediapipe/tasks/python/core:optional_dependencies", + "//mediapipe/tasks/python/core:task_info", + "//mediapipe/tasks/python/vision/core:base_vision_task_api", + "//mediapipe/tasks/python/vision/core:image_processing_options", + "//mediapipe/tasks/python/vision/core:vision_task_running_mode", + ], +) + py_library( name = "face_stylizer", srcs = [ diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py new file mode 100644 index 0000000000..a1877ff670 --- /dev/null +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -0,0 +1,567 @@ +# Copyright 2022 The MediaPipe Authors. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +"""MediaPipe holistic landmarker task.""" + +import dataclasses +from typing import Callable, Mapping, Optional, List + +from mediapipe.framework.formats import classification_pb2 +from mediapipe.framework.formats import landmark_pb2 +from mediapipe.python import packet_creator +from mediapipe.python import packet_getter +from mediapipe.python._framework_bindings import image as image_module +from mediapipe.python._framework_bindings import packet as packet_module +from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_landmarker_graph_options_pb2 +from mediapipe.tasks.python.components.containers import category as category_module +from mediapipe.tasks.python.components.containers import landmark as landmark_module +from mediapipe.tasks.python.core import base_options as base_options_module +from mediapipe.tasks.python.core import task_info as task_info_module +from mediapipe.tasks.python.core.optional_dependencies import doc_controls +from mediapipe.tasks.python.vision.core import base_vision_task_api +from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module +from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module + +_BaseOptions = base_options_module.BaseOptions +_HolisticLandmarkerGraphOptionsProto = ( + holistic_landmarker_graph_options_pb2.HolisticLandmarkerGraphOptions +) +_RunningMode = running_mode_module.VisionTaskRunningMode +_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions +_TaskInfo = task_info_module.TaskInfo + +_IMAGE_IN_STREAM_NAME = 'image_in' +_IMAGE_OUT_STREAM_NAME = 'image_out' +_IMAGE_TAG = 'IMAGE' +_NORM_RECT_STREAM_NAME = 'norm_rect_in' +_NORM_RECT_TAG = 'NORM_RECT' + + +_POSE_LANDMARKS_STREAM_NAME = "pose_landmarks" +_POSE_LANDMARKS_TAG_NAME = "POSE_LANDMARKS" +_POSE_WORLD_LANDMARKS_STREAM_NAME = "pose_world_landmarks" +_POSE_WORLD_LANDMARKS_TAG = "POSE_WORLD_LANDMARKS" +_POSE_SEGMENTATION_MASK_STREAM_NAME = "pose_segmentation_mask" +_POSE_SEGMENTATION_MASK_TAG = "pose_segmentation_mask" +_FACE_LANDMARKS_STREAM_NAME = "face_landmarks" +_FACE_LANDMARKS_TAG = "FACE_LANDMARKS" +_FACE_BLENDSHAPES_STREAM_NAME = "extra_blendshapes" +_FACE_BLENDSHAPES_TAG = "FACE_BLENDSHAPES" +_LEFT_HAND_LANDMARKS_STREAM_NAME = "left_hand_landmarks" +_LEFT_HAND_LANDMARKS_TAG = "LEFT_HAND_LANDMARKS" +_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME = "left_hand_world_landmarks" +_LEFT_HAND_WORLD_LANDMARKS_TAG = "LEFT_HAND_WORLD_LANDMARKS" +_RIGHT_HAND_LANDMARKS_STREAM_NAME = "right_hand_landmarks" +_RIGHT_HAND_LANDMARKS_TAG = "RIGHT_HAND_LANDMARKS" +_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME = "right_hand_world_landmarks" +_RIGHT_HAND_WORLD_LANDMARKS_TAG = "RIGHT_HAND_WORLD_LANDMARKS" + +_TASK_GRAPH_NAME = 'mediapipe.tasks.vision.holistic_landmarker.HolisticLandmarkerGraph' +_MICRO_SECONDS_PER_MILLISECOND = 1000 + + +@dataclasses.dataclass +class HolisticLandmarkerResult: + """The holistic landmarks result from HolisticLandmarker, where each vector element represents a single holistic detected in the image. + + Attributes: + TODO + """ + face_landmarks: List[List[landmark_module.NormalizedLandmark]] + pose_landmarks: List[List[landmark_module.NormalizedLandmark]] + pose_world_landmarks: List[List[landmark_module.Landmark]] + left_hand_landmarks: List[List[landmark_module.NormalizedLandmark]] + left_hand_world_landmarks: List[List[landmark_module.Landmark]] + right_hand_landmarks: List[List[landmark_module.NormalizedLandmark]] + right_hand_world_landmarks: List[List[landmark_module.Landmark]] + face_blendshapes: Optional[List[List[category_module.Category]]] = None + segmentation_masks: Optional[List[image_module.Image]] = None + + +def _build_landmarker_result( + output_packets: Mapping[str, packet_module.Packet] +) -> HolisticLandmarkerResult: + """Constructs a `HolisticLandmarksDetectionResult` from output packets.""" + holistic_landmarker_result = HolisticLandmarkerResult([], [], [], [], [], [], + []) + + face_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_FACE_LANDMARKS_STREAM_NAME] + ) + + if _POSE_SEGMENTATION_MASK_STREAM_NAME in output_packets: + holistic_landmarker_result.segmentation_masks = packet_getter.get_image_list( + output_packets[_POSE_SEGMENTATION_MASK_STREAM_NAME] + ) + + pose_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_POSE_LANDMARKS_STREAM_NAME] + ) + + pose_world_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_POSE_WORLD_LANDMARKS_STREAM_NAME] + ) + + left_hand_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_LEFT_HAND_LANDMARKS_STREAM_NAME] + ) + + left_hand_world_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] + ) + + right_hand_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_RIGHT_HAND_LANDMARKS_STREAM_NAME] + ) + + right_hand_world_landmarks_proto_list = packet_getter.get_proto_list( + output_packets[_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] + ) + + face_landmarks_results = [] + for proto in face_landmarks_proto_list: + face_landmarks = landmark_pb2.NormalizedLandmarkList() + face_landmarks.MergeFrom(proto) + face_landmarks_list = [] + for face_landmark in face_landmarks.landmark: + face_landmarks_list.append( + landmark_module.NormalizedLandmark.create_from_pb2(face_landmark) + ) + face_landmarks_results.append(face_landmarks_list) + + face_blendshapes_results = [] + if _FACE_BLENDSHAPES_STREAM_NAME in output_packets: + face_blendshapes_proto_list = packet_getter.get_proto_list( + output_packets[_FACE_BLENDSHAPES_STREAM_NAME] + ) + for proto in face_blendshapes_proto_list: + face_blendshapes_categories = [] + face_blendshapes_classifications = classification_pb2.ClassificationList() + face_blendshapes_classifications.MergeFrom(proto) + for face_blendshapes in face_blendshapes_classifications.classification: + face_blendshapes_categories.append( + category_module.Category( + index=face_blendshapes.index, + score=face_blendshapes.score, + display_name=face_blendshapes.display_name, + category_name=face_blendshapes.label, + ) + ) + face_blendshapes_results.append(face_blendshapes_categories) + + for proto in pose_landmarks_proto_list: + pose_landmarks = landmark_pb2.NormalizedLandmarkList() + pose_landmarks.MergeFrom(proto) + pose_landmarks_list = [] + for pose_landmark in pose_landmarks.landmark: + pose_landmarks_list.append( + landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark) + ) + holistic_landmarker_result.pose_landmarks.append(pose_landmarks_list) + + for proto in pose_world_landmarks_proto_list: + pose_world_landmarks = landmark_pb2.LandmarkList() + pose_world_landmarks.MergeFrom(proto) + pose_world_landmarks_list = [] + for pose_world_landmark in pose_world_landmarks.landmark: + pose_world_landmarks_list.append( + landmark_module.Landmark.create_from_pb2(pose_world_landmark) + ) + holistic_landmarker_result.pose_world_landmarks.append( + pose_world_landmarks_list + ) + + for proto in left_hand_landmarks_proto_list: + left_hand_landmarks = landmark_pb2.NormalizedLandmarkList() + left_hand_landmarks.MergeFrom(proto) + left_hand_landmarks_list = [] + for hand_landmark in left_hand_landmarks.landmark: + left_hand_landmarks_list.append( + landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) + ) + holistic_landmarker_result.left_hand_landmarks.append( + left_hand_landmarks_list + ) + + for proto in left_hand_world_landmarks_proto_list: + left_hand_world_landmarks = landmark_pb2.LandmarkList() + left_hand_world_landmarks.MergeFrom(proto) + left_hand_world_landmarks_list = [] + for left_hand_world_landmark in left_hand_world_landmarks.landmark: + left_hand_world_landmarks_list.append( + landmark_module.Landmark.create_from_pb2(left_hand_world_landmark) + ) + holistic_landmarker_result.left_hand_world_landmarks.append( + left_hand_world_landmarks_list + ) + + for proto in right_hand_landmarks_proto_list: + right_hand_landmarks = landmark_pb2.NormalizedLandmarkList() + right_hand_landmarks.MergeFrom(proto) + right_hand_landmarks_list = [] + for hand_landmark in right_hand_landmarks.landmark: + right_hand_landmarks_list.append( + landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) + ) + holistic_landmarker_result.right_hand_landmarks.append( + right_hand_landmarks_list + ) + + for proto in right_hand_world_landmarks_proto_list: + right_hand_world_landmarks = landmark_pb2.LandmarkList() + right_hand_world_landmarks.MergeFrom(proto) + right_hand_world_landmarks_list = [] + for right_hand_world_landmark in right_hand_world_landmarks.landmark: + right_hand_world_landmarks_list.append( + landmark_module.Landmark.create_from_pb2(right_hand_world_landmark) + ) + holistic_landmarker_result.right_hand_world_landmarks.append( + right_hand_world_landmarks_list + ) + + return holistic_landmarker_result + + +@dataclasses.dataclass +class HolisticLandmarkerOptions: + """Options for the holistic landmarker task. + + Attributes: + base_options: Base options for the holistic landmarker task. + running_mode: The running mode of the task. Default to the image mode. + HolisticLandmarker has three running modes: 1) The image mode for + detecting holistic landmarks on single image inputs. 2) The video mode for + detecting holistic landmarks on the decoded frames of a video. 3) The live + stream mode for detecting holistic landmarks on the live stream of input + data, such as from camera. In this mode, the "result_callback" below must + be specified to receive the detection results asynchronously. + min_face_detection_confidence: The minimum confidence score for the face + detection to be considered successful. + min_face_suppression_threshold: The minimum non-maximum-suppression + threshold for face detection to be considered overlapped. + min_face_landmarks_confidence: The minimum confidence score for the face + landmark detection to be considered successful. + min_pose_detection_confidence: The minimum confidence score for the pose + detection to be considered successful. + min_pose_suppression_threshold: The minimum non-maximum-suppression + threshold for pose detection to be considered overlapped. + min_pose_landmarks_confidence: The minimum confidence score for the pose + landmark detection to be considered successful. + min_hand_landmarks_confidence: The minimum confidence score for the hand + landmark detection to be considered successful. + result_callback: The user-defined result callback for processing live stream + data. The result callback should only be specified when the running mode + is set to the live stream mode. + """ + + base_options: _BaseOptions + running_mode: _RunningMode = _RunningMode.IMAGE + num_holistics: int = 1 + min_face_detection_confidence: float = 0.5 + min_face_suppression_threshold: float = 0.5 + min_face_landmarks_confidence: float = 0.5 + min_pose_detection_confidence: float = 0.5 + min_pose_suppression_threshold: float = 0.5 + min_pose_landmarks_confidence: float = 0.5 + min_hand_landmarks_confidence: float = 0.5 + output_face_blendshapes: bool = False + output_segmentation_masks: bool = False + result_callback: Optional[ + Callable[[HolisticLandmarkerResult, image_module.Image, int], None] + ] = None + + @doc_controls.do_not_generate_docs + def to_pb2(self) -> _HolisticLandmarkerGraphOptionsProto: + """Generates an HolisticLandmarkerGraphOptions protobuf object.""" + base_options_proto = self.base_options.to_pb2() + base_options_proto.use_stream_mode = ( + False if self.running_mode == _RunningMode.IMAGE else True + ) + + # Initialize the holistic landmarker options from base options. + holistic_landmarker_options_proto = _HolisticLandmarkerGraphOptionsProto( + base_options=base_options_proto + ) + # Configure face detector and face landmarks detector options. + # holistic_landmarker_options_proto.face_detector_graph_options.min_detection_confidence = ( + # self.min_face_detection_confidence + # ) + # holistic_landmarker_options_proto.face_detector_graph_options.min_suppression_threshold = ( + # self.min_face_suppression_threshold + # ) + # holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( + # self.min_face_landmarks_confidence + # ) + # # Configure pose detector and pose landmarks detector options. + # holistic_landmarker_options_proto.pose_detector_graph_options.min_detection_confidence = ( + # self.min_pose_detection_confidence + # ) + # holistic_landmarker_options_proto.pose_detector_graph_options.min_suppression_threshold = ( + # self.min_pose_suppression_threshold + # ) + # holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( + # self.min_pose_landmarks_confidence + # ) + # # Configure hand landmarks detector options. + # holistic_landmarker_options_proto.hand_landmarks_detector_graph_options.min_detection_confidence = ( + # self.min_hand_landmarks_confidence + # ) + return holistic_landmarker_options_proto + + +class HolisticLandmarker(base_vision_task_api.BaseVisionTaskApi): + """Class that performs holistic landmarks detection on images.""" + + @classmethod + def create_from_model_path(cls, model_path: str) -> 'HolisticLandmarker': + """Creates an `HolisticLandmarker` object from a TensorFlow Lite model and the default `HolisticLandmarkerOptions`. + + Note that the created `HolisticLandmarker` instance is in image mode, for + detecting holistic landmarks on single image inputs. + + Args: + model_path: Path to the model. + + Returns: + `HolisticLandmarker` object that's created from the model file and the + default `HolisticLandmarkerOptions`. + + Raises: + ValueError: If failed to create `HolisticLandmarker` object from the + provided file such as invalid file path. + RuntimeError: If other types of error occurred. + """ + base_options = _BaseOptions(model_asset_path=model_path) + options = HolisticLandmarkerOptions( + base_options=base_options, running_mode=_RunningMode.IMAGE + ) + return cls.create_from_options(options) + + @classmethod + def create_from_options( + cls, options: HolisticLandmarkerOptions + ) -> 'HolisticLandmarker': + """Creates the `HolisticLandmarker` object from holistic landmarker options. + + Args: + options: Options for the holistic landmarker task. + + Returns: + `HolisticLandmarker` object that's created from `options`. + + Raises: + ValueError: If failed to create `HolisticLandmarker` object from + `HolisticLandmarkerOptions` such as missing the model. + RuntimeError: If other types of error occurred. + """ + + def packets_callback(output_packets: Mapping[str, packet_module.Packet]): + if output_packets[_IMAGE_OUT_STREAM_NAME].is_empty(): + return + + image = packet_getter.get_image(output_packets[_IMAGE_OUT_STREAM_NAME]) + + if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty(): + empty_packet = output_packets[_FACE_LANDMARKS_STREAM_NAME] + options.result_callback( + HolisticLandmarkerResult([], [], [], [], [], [], []), + image, + empty_packet.timestamp.value // _MICRO_SECONDS_PER_MILLISECOND, + ) + return + + holistic_landmarks_detection_result = _build_landmarker_result(output_packets) + timestamp = output_packets[_FACE_LANDMARKS_STREAM_NAME].timestamp + options.result_callback( + holistic_landmarks_detection_result, + image, + timestamp.value // _MICRO_SECONDS_PER_MILLISECOND, + ) + + output_streams = [ + ':'.join([_FACE_LANDMARKS_TAG, _FACE_LANDMARKS_STREAM_NAME]), + ':'.join([_POSE_LANDMARKS_TAG_NAME, _POSE_LANDMARKS_STREAM_NAME]), + ':'.join( + [_POSE_WORLD_LANDMARKS_TAG, _POSE_WORLD_LANDMARKS_STREAM_NAME] + ), + ':'.join([_LEFT_HAND_LANDMARKS_TAG, _LEFT_HAND_LANDMARKS_STREAM_NAME]), + ':'.join( + [_LEFT_HAND_WORLD_LANDMARKS_TAG, _LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] + ), + ':'.join([_RIGHT_HAND_LANDMARKS_TAG, _RIGHT_HAND_LANDMARKS_STREAM_NAME]), + ':'.join( + [_RIGHT_HAND_WORLD_LANDMARKS_TAG, _RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] + ), + ':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME]), + ] + + if options.output_segmentation_masks: + output_streams.append( + ':'.join([_POSE_SEGMENTATION_MASK_TAG, _POSE_SEGMENTATION_MASK_STREAM_NAME]) + ) + + if options.output_face_blendshapes: + output_streams.append( + ':'.join([_FACE_BLENDSHAPES_TAG, _FACE_BLENDSHAPES_STREAM_NAME]) + ) + + task_info = _TaskInfo( + task_graph=_TASK_GRAPH_NAME, + input_streams=[ + ':'.join([_IMAGE_TAG, _IMAGE_IN_STREAM_NAME]), + ':'.join([_NORM_RECT_TAG, _NORM_RECT_STREAM_NAME]), + ], + output_streams=output_streams, + task_options=options, + ) + return cls( + task_info.generate_graph_config( + enable_flow_limiting=options.running_mode + == _RunningMode.LIVE_STREAM + ), + options.running_mode, + packets_callback if options.result_callback else None, + ) + + def detect( + self, + image: image_module.Image, + image_processing_options: Optional[_ImageProcessingOptions] = None, + ) -> HolisticLandmarkerResult: + """Performs holistic landmarks detection on the given image. + + Only use this method when the HolisticLandmarker is created with the image + running mode. + + The image can be of any size with format RGB or RGBA. + TODO: Describes how the input image will be preprocessed after the yuv + support is implemented. + + Args: + image: MediaPipe Image. + image_processing_options: Options for image processing. + + Returns: + The holistic landmarks detection results. + + Raises: + ValueError: If any of the input arguments is invalid. + RuntimeError: If holistic landmarker detection failed to run. + """ + normalized_rect = self.convert_to_normalized_rect( + image_processing_options, image, roi_allowed=False + ) + output_packets = self._process_image_data({ + _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image), + _NORM_RECT_STREAM_NAME: packet_creator.create_proto( + normalized_rect.to_pb2() + ), + }) + + if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty(): + return HolisticLandmarkerResult([], [], [], [], [], [], []) + + return _build_landmarker_result(output_packets) + + def detect_for_video( + self, + image: image_module.Image, + timestamp_ms: int, + image_processing_options: Optional[_ImageProcessingOptions] = None, + ) -> HolisticLandmarkerResult: + """Performs holistic landmarks detection on the provided video frame. + + Only use this method when the HolisticLandmarker is created with the video + running mode. + + Only use this method when the HolisticLandmarker is created with the video + running mode. It's required to provide the video frame's timestamp (in + milliseconds) along with the video frame. The input timestamps should be + monotonically increasing for adjacent calls of this method. + + Args: + image: MediaPipe Image. + timestamp_ms: The timestamp of the input video frame in milliseconds. + image_processing_options: Options for image processing. + + Returns: + The holistic landmarks detection results. + + Raises: + ValueError: If any of the input arguments is invalid. + RuntimeError: If holistic landmarker detection failed to run. + """ + normalized_rect = self.convert_to_normalized_rect( + image_processing_options, image, roi_allowed=False + ) + output_packets = self._process_video_data({ + _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at( + timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND + ), + _NORM_RECT_STREAM_NAME: packet_creator.create_proto( + normalized_rect.to_pb2() + ).at(timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND), + }) + + if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty(): + return HolisticLandmarkerResult([], [], [], [], [], [], []) + + return _build_landmarker_result(output_packets) + + def detect_async( + self, + image: image_module.Image, + timestamp_ms: int, + image_processing_options: Optional[_ImageProcessingOptions] = None, + ) -> None: + """Sends live image data to perform holistic landmarks detection. + + The results will be available via the "result_callback" provided in the + HolisticLandmarkerOptions. Only use this method when the HolisticLandmarker is + created with the live stream running mode. + + Only use this method when the HolisticLandmarker is created with the live + stream running mode. The input timestamps should be monotonically increasing + for adjacent calls of this method. This method will return immediately after + the input image is accepted. The results will be available via the + `result_callback` provided in the `HolisticLandmarkerOptions`. The + `detect_async` method is designed to process live stream data such as + camera input. To lower the overall latency, holistic landmarker may drop the + input images if needed. In other words, it's not guaranteed to have output + per input image. + + The `result_callback` provides: + - The holistic landmarks detection results. + - The input image that the holistic landmarker runs on. + - The input timestamp in milliseconds. + + Args: + image: MediaPipe Image. + timestamp_ms: The timestamp of the input image in milliseconds. + image_processing_options: Options for image processing. + + Raises: + ValueError: If the current input timestamp is smaller than what the + holistic landmarker has already processed. + """ + normalized_rect = self.convert_to_normalized_rect( + image_processing_options, image, roi_allowed=False + ) + self._send_live_stream_data({ + _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at( + timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND + ), + _NORM_RECT_STREAM_NAME: packet_creator.create_proto( + normalized_rect.to_pb2() + ).at(timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND), + }) diff --git a/mediapipe/tasks/testdata/vision/BUILD b/mediapipe/tasks/testdata/vision/BUILD index 2f51573091..4220e29bb9 100644 --- a/mediapipe/tasks/testdata/vision/BUILD +++ b/mediapipe/tasks/testdata/vision/BUILD @@ -57,9 +57,11 @@ mediapipe_files(srcs = [ "hand_landmark_lite.tflite", "hand_landmarker.task", "handrecrop_2020_07_21_v0.f16.tflite", + "holistic_landmarker.task", "left_hands.jpg", "left_hands_rotated.jpg", "leopard_bg_removal_result_512x512.png", + "male_full_height_hands.jpg", "mobilenet_v1_0.25_192_quantized_1_default_1.tflite", "mobilenet_v1_0.25_224_1_default_1.tflite", "mobilenet_v1_0.25_224_1_metadata_1.tflite", @@ -138,9 +140,11 @@ filegroup( "fist.png", "hand_landmark_full.tflite", "hand_landmark_lite.tflite", + "holistic_landmarker.task", "left_hands.jpg", "left_hands_rotated.jpg", "leopard_bg_removal_result_512x512.png", + "male_full_height_hands.jpg", "mozart_square.jpg", "multi_objects.jpg", "multi_objects_rotated.jpg", From ea95ae753df650245b3747017d18285a8ac49cbc Mon Sep 17 00:00:00 2001 From: Kinar Date: Mon, 4 Dec 2023 04:12:22 -0800 Subject: [PATCH 2/9] Uncommented configuring options for Holistic Landmarker --- .../python/vision/holistic_landmarker.py | 46 +++++++++---------- 1 file changed, 23 insertions(+), 23 deletions(-) diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py index a1877ff670..940663230a 100644 --- a/mediapipe/tasks/python/vision/holistic_landmarker.py +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -293,29 +293,29 @@ def to_pb2(self) -> _HolisticLandmarkerGraphOptionsProto: base_options=base_options_proto ) # Configure face detector and face landmarks detector options. - # holistic_landmarker_options_proto.face_detector_graph_options.min_detection_confidence = ( - # self.min_face_detection_confidence - # ) - # holistic_landmarker_options_proto.face_detector_graph_options.min_suppression_threshold = ( - # self.min_face_suppression_threshold - # ) - # holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( - # self.min_face_landmarks_confidence - # ) - # # Configure pose detector and pose landmarks detector options. - # holistic_landmarker_options_proto.pose_detector_graph_options.min_detection_confidence = ( - # self.min_pose_detection_confidence - # ) - # holistic_landmarker_options_proto.pose_detector_graph_options.min_suppression_threshold = ( - # self.min_pose_suppression_threshold - # ) - # holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( - # self.min_pose_landmarks_confidence - # ) - # # Configure hand landmarks detector options. - # holistic_landmarker_options_proto.hand_landmarks_detector_graph_options.min_detection_confidence = ( - # self.min_hand_landmarks_confidence - # ) + holistic_landmarker_options_proto.face_detector_graph_options.min_detection_confidence = ( + self.min_face_detection_confidence + ) + holistic_landmarker_options_proto.face_detector_graph_options.min_suppression_threshold = ( + self.min_face_suppression_threshold + ) + holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( + self.min_face_landmarks_confidence + ) + # Configure pose detector and pose landmarks detector options. + holistic_landmarker_options_proto.pose_detector_graph_options.min_detection_confidence = ( + self.min_pose_detection_confidence + ) + holistic_landmarker_options_proto.pose_detector_graph_options.min_suppression_threshold = ( + self.min_pose_suppression_threshold + ) + holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( + self.min_pose_landmarks_confidence + ) + # Configure hand landmarks detector options. + holistic_landmarker_options_proto.hand_landmarks_detector_graph_options.min_detection_confidence = ( + self.min_hand_landmarks_confidence + ) return holistic_landmarker_options_proto From 24fe8eb73ab0792c4ec22b28c4fe8cdfae58255b Mon Sep 17 00:00:00 2001 From: Kinar Date: Sun, 17 Dec 2023 15:13:34 -0800 Subject: [PATCH 3/9] Support both proto2 and proto3 in task subgraph options configuration, and revised the Holistic Landmarker API's implementation --- mediapipe/tasks/python/core/BUILD | 1 + mediapipe/tasks/python/core/task_info.py | 30 ++- mediapipe/tasks/python/test/vision/BUILD | 1 + .../test/vision/holistic_landmarker_test.py | 101 ++++++- mediapipe/tasks/python/vision/BUILD | 1 + .../python/vision/holistic_landmarker.py | 254 +++++++++--------- 6 files changed, 235 insertions(+), 153 deletions(-) diff --git a/mediapipe/tasks/python/core/BUILD b/mediapipe/tasks/python/core/BUILD index 9d2dc3f0bd..f256430c3b 100644 --- a/mediapipe/tasks/python/core/BUILD +++ b/mediapipe/tasks/python/core/BUILD @@ -49,5 +49,6 @@ py_library( "//mediapipe/calculators/core:flow_limiter_calculator_py_pb2", "//mediapipe/framework:calculator_options_py_pb2", "//mediapipe/framework:calculator_py_pb2", + "@com_google_protobuf//:protobuf_python" ], ) diff --git a/mediapipe/tasks/python/core/task_info.py b/mediapipe/tasks/python/core/task_info.py index 8941033613..5d039a0341 100644 --- a/mediapipe/tasks/python/core/task_info.py +++ b/mediapipe/tasks/python/core/task_info.py @@ -21,6 +21,7 @@ from mediapipe.framework import calculator_options_pb2 from mediapipe.framework import calculator_pb2 from mediapipe.tasks.python.core.optional_dependencies import doc_controls +from google.protobuf.any_pb2 import Any @doc_controls.do_not_generate_docs @@ -80,22 +81,31 @@ def add_stream_name_prefix(tag_index_name): raise ValueError( '`task_options` doesn`t provide `to_pb2()` method to convert itself to be a protobuf object.' ) - task_subgraph_options = calculator_options_pb2.CalculatorOptions() + task_options_proto = self.task_options.to_pb2() - # For protobuf 2 compat. + node_config = calculator_pb2.CalculatorGraphConfig.Node( + calculator=self.task_graph, + input_stream=self.input_streams, + output_stream=self.output_streams + ) + if hasattr(task_options_proto, 'ext'): + # Use the extension mechanism for task_subgraph_options (proto2) + task_subgraph_options = calculator_options_pb2.CalculatorOptions() task_subgraph_options.Extensions[task_options_proto.ext].CopyFrom( task_options_proto) + node_config.options.CopyFrom(task_subgraph_options) + else: + # Use the Any type for task_subgraph_options (proto3) + task_subgraph_options = Any() + task_subgraph_options.Pack(self.task_options.to_pb2()) + node_config.node_options.append(task_subgraph_options) if not enable_flow_limiting: return calculator_pb2.CalculatorGraphConfig( node=[ - calculator_pb2.CalculatorGraphConfig.Node( - calculator=self.task_graph, - input_stream=self.input_streams, - output_stream=self.output_streams, - options=task_subgraph_options) + node_config ], input_stream=self.input_streams, output_stream=self.output_streams) @@ -125,11 +135,7 @@ def add_stream_name_prefix(tag_index_name): options=flow_limiter_options) config = calculator_pb2.CalculatorGraphConfig( node=[ - calculator_pb2.CalculatorGraphConfig.Node( - calculator=self.task_graph, - input_stream=task_subgraph_inputs, - output_stream=self.output_streams, - options=task_subgraph_options), flow_limiter + node_config, flow_limiter ], input_stream=self.input_streams, output_stream=self.output_streams) diff --git a/mediapipe/tasks/python/test/vision/BUILD b/mediapipe/tasks/python/test/vision/BUILD index 374ba689c1..eae05de4d9 100644 --- a/mediapipe/tasks/python/test/vision/BUILD +++ b/mediapipe/tasks/python/test/vision/BUILD @@ -206,6 +206,7 @@ py_test( deps = [ "//mediapipe/framework/formats:classification_py_pb2", "//mediapipe/framework/formats:landmark_py_pb2", + "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_result_py_pb2", "//mediapipe/python:_framework_bindings", "//mediapipe/tasks/python/components/containers:category", "//mediapipe/tasks/python/components/containers:landmark", diff --git a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py index 0c91793013..6f27c903db 100644 --- a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py +++ b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py @@ -14,6 +14,7 @@ """Tests for holistic landmarker.""" import enum +from typing import List from unittest import mock from absl.testing import absltest @@ -23,6 +24,7 @@ from google.protobuf import text_format from mediapipe.framework.formats import classification_pb2 from mediapipe.framework.formats import landmark_pb2 +from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_result_pb2 from mediapipe.python._framework_bindings import image as image_module from mediapipe.tasks.python.components.containers import category as category_module from mediapipe.tasks.python.components.containers import landmark as landmark_module @@ -35,6 +37,7 @@ HolisticLandmarkerResult = holistic_landmarker.HolisticLandmarkerResult +_HolisticResultProto = holistic_result_pb2.HolisticResult _BaseOptions = base_options_module.BaseOptions _Category = category_module.Category _Rect = rect_module.Rect @@ -46,14 +49,31 @@ _RUNNING_MODE = running_mode_module.VisionTaskRunningMode _ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions -_HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE = 'face_landmarker.task' +_HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE = 'holistic_landmarker.task' _POSE_IMAGE = 'male_full_height_hands.jpg' _CAT_IMAGE = 'cat.jpg' -_HOLISTIC_RESULT = "male_full_height_hands_result_cpu.pbtxt" +_EXPECTED_HOLISTIC_RESULT = "male_full_height_hands_result_cpu.pbtxt" _LANDMARKS_MARGIN = 0.03 _BLENDSHAPES_MARGIN = 0.13 +def _get_expected_holistic_landmarker_result( + file_path: str, +) -> HolisticLandmarkerResult: + holistic_result_file_path = test_utils.get_test_data_path( + file_path + ) + with open(holistic_result_file_path, 'rb') as f: + holistic_result_proto = _HolisticResultProto() + # Use this if a .pb file is available. + # holistic_result_proto.ParseFromString(f.read()) + text_format.Parse(f.read(), holistic_result_proto) + holistic_landmarker_result = HolisticLandmarkerResult.create_from_pb2( + holistic_result_proto + ) + return holistic_landmarker_result + + class ModelFileType(enum.Enum): FILE_CONTENT = 1 FILE_NAME = 2 @@ -70,20 +90,77 @@ def setUp(self): _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE ) + def _expect_landmarks_correct( + self, actual_landmarks, expected_landmarks, margin + ): + # Expects to have the same number of poses detected. + self.assertLen(actual_landmarks, len(expected_landmarks)) + + for i, elem in enumerate(actual_landmarks): + self.assertAlmostEqual(elem.x, expected_landmarks[i].x, delta=margin) + self.assertAlmostEqual(elem.y, expected_landmarks[i].y, delta=margin) + + def _expect_blendshapes_correct( + self, actual_blendshapes, expected_blendshapes, margin + ): + # Expects to have the same number of blendshapes. + self.assertLen(actual_blendshapes, len(expected_blendshapes)) + + for i, elem in enumerate(actual_blendshapes): + self.assertEqual(elem.index, expected_blendshapes[i].index) + self.assertAlmostEqual( + elem.score, + expected_blendshapes[i].score, + delta=margin, + ) + + def _expect_holistic_landmarker_results_correct( + self, + actual_result: HolisticLandmarkerResult, + expected_result: HolisticLandmarkerResult, + output_segmentation_masks: bool, + landmarks_margin: float, + blendshapes_margin: float, + ): + self._expect_landmarks_correct( + actual_result.pose_landmarks, expected_result.pose_landmarks, + landmarks_margin + ) + self._expect_landmarks_correct( + actual_result.face_landmarks, expected_result.face_landmarks, + landmarks_margin + ) + self._expect_blendshapes_correct( + actual_result.face_blendshapes, expected_result.face_blendshapes, + blendshapes_margin + ) + if output_segmentation_masks: + self.assertIsInstance(actual_result.segmentation_masks, List) + for _, mask in enumerate(actual_result.segmentation_masks): + self.assertIsInstance(mask, _Image) + else: + self.assertIsNone(actual_result.segmentation_masks) + @parameterized.parameters( ( ModelFileType.FILE_NAME, - _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + False, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) ), ( ModelFileType.FILE_CONTENT, - _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + False, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) ), ) def test_detect( self, model_file_type, - model_name + model_name, + output_segmentation_masks, + expected_holistic_landmarker_result: HolisticLandmarkerResult ): # Creates holistic landmarker. model_path = test_utils.get_test_data_path(model_name) @@ -98,15 +175,21 @@ def test_detect( raise ValueError('model_file_type is invalid.') options = _HolisticLandmarkerOptions( - base_options=base_options + base_options=base_options, + output_face_blendshapes=True + if expected_holistic_landmarker_result.face_blendshapes else False, + output_segmentation_masks=output_segmentation_masks, ) landmarker = _HolisticLandmarker.create_from_options(options) # Performs holistic landmarks detection on the input. detection_result = landmarker.detect(self.test_image) - - # Closes the holistic landmarker explicitly when the holistic landmarker is not used - # in a context. + self._expect_holistic_landmarker_results_correct( + detection_result, expected_holistic_landmarker_result, + output_segmentation_masks, _LANDMARKS_MARGIN, _BLENDSHAPES_MARGIN + ) + # Closes the holistic landmarker explicitly when the holistic landmarker is + # not used in a context. landmarker.close() diff --git a/mediapipe/tasks/python/vision/BUILD b/mediapipe/tasks/python/vision/BUILD index 8253a92327..1b0a1454bf 100644 --- a/mediapipe/tasks/python/vision/BUILD +++ b/mediapipe/tasks/python/vision/BUILD @@ -254,6 +254,7 @@ py_library( "//mediapipe/python:_framework_bindings", "//mediapipe/python:packet_creator", "//mediapipe/python:packet_getter", + "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_result_py_pb2", "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_landmarker_graph_options_py_pb2", "//mediapipe/tasks/python/components/containers:category", "//mediapipe/tasks/python/components/containers:landmark", diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py index 940663230a..8edfaf445d 100644 --- a/mediapipe/tasks/python/vision/holistic_landmarker.py +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -22,6 +22,7 @@ from mediapipe.python import packet_getter from mediapipe.python._framework_bindings import image as image_module from mediapipe.python._framework_bindings import packet as packet_module +from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_result_pb2 from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_landmarker_graph_options_pb2 from mediapipe.tasks.python.components.containers import category as category_module from mediapipe.tasks.python.components.containers import landmark as landmark_module @@ -33,6 +34,7 @@ from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module _BaseOptions = base_options_module.BaseOptions +_HolisticResultProto = holistic_result_pb2.HolisticResult _HolisticLandmarkerGraphOptionsProto = ( holistic_landmarker_graph_options_pb2.HolisticLandmarkerGraphOptions ) @@ -43,9 +45,6 @@ _IMAGE_IN_STREAM_NAME = 'image_in' _IMAGE_OUT_STREAM_NAME = 'image_out' _IMAGE_TAG = 'IMAGE' -_NORM_RECT_STREAM_NAME = 'norm_rect_in' -_NORM_RECT_TAG = 'NORM_RECT' - _POSE_LANDMARKS_STREAM_NAME = "pose_landmarks" _POSE_LANDMARKS_TAG_NAME = "POSE_LANDMARKS" @@ -77,16 +76,64 @@ class HolisticLandmarkerResult: Attributes: TODO """ - face_landmarks: List[List[landmark_module.NormalizedLandmark]] - pose_landmarks: List[List[landmark_module.NormalizedLandmark]] - pose_world_landmarks: List[List[landmark_module.Landmark]] - left_hand_landmarks: List[List[landmark_module.NormalizedLandmark]] - left_hand_world_landmarks: List[List[landmark_module.Landmark]] - right_hand_landmarks: List[List[landmark_module.NormalizedLandmark]] - right_hand_world_landmarks: List[List[landmark_module.Landmark]] - face_blendshapes: Optional[List[List[category_module.Category]]] = None + face_landmarks: List[landmark_module.NormalizedLandmark] + pose_landmarks: List[landmark_module.NormalizedLandmark] + pose_world_landmarks:List[landmark_module.Landmark] + left_hand_landmarks: List[landmark_module.NormalizedLandmark] + left_hand_world_landmarks: List[landmark_module.Landmark] + right_hand_landmarks: List[landmark_module.NormalizedLandmark] + right_hand_world_landmarks: List[landmark_module.Landmark] + face_blendshapes: Optional[List[category_module.Category]] = None segmentation_masks: Optional[List[image_module.Image]] = None + @classmethod + @doc_controls.do_not_generate_docs + def create_from_pb2( + cls, + pb2_obj: _HolisticResultProto + ) -> 'HolisticLandmarkerResult': + """Creates a `HolisticLandmarkerResult` object from the given protobuf + object.""" + return HolisticLandmarkerResult( + face_landmarks=[ + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.face_landmarks.landmark + ] if hasattr(pb2_obj, 'face_landmarks') else None, + pose_landmarks=[ + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.pose_landmarks.landmark + ] if hasattr(pb2_obj, 'pose_landmarks') else None, + pose_world_landmarks=[ + landmark_module.Landmark.create_from_pb2(landmark) + for landmark in pb2_obj.pose_world_landmarks.landmark + ] if hasattr(pb2_obj, 'pose_world_landmarks') else None, + left_hand_landmarks=[ + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.left_hand_landmarks.landmark + ] if hasattr(pb2_obj, 'left_hand_landmarks') else None, + left_hand_world_landmarks=[ + landmark_module.Landmark.create_from_pb2(landmark) + for landmark in pb2_obj.left_hand_world_landmarks.landmark + ] if hasattr(pb2_obj, 'left_hand_world_landmarks') else None, + right_hand_landmarks=[ + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.right_hand_landmarks.landmark + ] if hasattr(pb2_obj, 'right_hand_landmarks') else None, + right_hand_world_landmarks=[ + landmark_module.Landmark.create_from_pb2(landmark) + for landmark in pb2_obj.right_hand_world_landmarks.landmark + ] if hasattr(pb2_obj, 'right_hand_world_landmarks') else None, + face_blendshapes=[ + category_module.Category( + score=classification.score, + index=classification.index, + category_name=classification.label, + display_name=classification.display_name + ) + for classification in pb2_obj.face_blendshapes.classification + ] if hasattr(pb2_obj, 'face_blendshapes') else None, + ) + def _build_landmarker_result( output_packets: Mapping[str, packet_module.Packet] @@ -95,138 +142,103 @@ def _build_landmarker_result( holistic_landmarker_result = HolisticLandmarkerResult([], [], [], [], [], [], []) - face_landmarks_proto_list = packet_getter.get_proto_list( + face_landmarks_proto_list = packet_getter.get_proto( output_packets[_FACE_LANDMARKS_STREAM_NAME] ) - if _POSE_SEGMENTATION_MASK_STREAM_NAME in output_packets: - holistic_landmarker_result.segmentation_masks = packet_getter.get_image_list( - output_packets[_POSE_SEGMENTATION_MASK_STREAM_NAME] - ) - - pose_landmarks_proto_list = packet_getter.get_proto_list( + pose_landmarks_proto_list = packet_getter.get_proto( output_packets[_POSE_LANDMARKS_STREAM_NAME] ) - pose_world_landmarks_proto_list = packet_getter.get_proto_list( + pose_world_landmarks_proto_list = packet_getter.get_proto( output_packets[_POSE_WORLD_LANDMARKS_STREAM_NAME] ) - left_hand_landmarks_proto_list = packet_getter.get_proto_list( + left_hand_landmarks_proto_list = packet_getter.get_proto( output_packets[_LEFT_HAND_LANDMARKS_STREAM_NAME] ) - left_hand_world_landmarks_proto_list = packet_getter.get_proto_list( + left_hand_world_landmarks_proto_list = packet_getter.get_proto( output_packets[_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] ) - right_hand_landmarks_proto_list = packet_getter.get_proto_list( + right_hand_landmarks_proto_list = packet_getter.get_proto( output_packets[_RIGHT_HAND_LANDMARKS_STREAM_NAME] ) - right_hand_world_landmarks_proto_list = packet_getter.get_proto_list( + right_hand_world_landmarks_proto_list = packet_getter.get_proto( output_packets[_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] ) - face_landmarks_results = [] - for proto in face_landmarks_proto_list: - face_landmarks = landmark_pb2.NormalizedLandmarkList() - face_landmarks.MergeFrom(proto) - face_landmarks_list = [] - for face_landmark in face_landmarks.landmark: - face_landmarks_list.append( - landmark_module.NormalizedLandmark.create_from_pb2(face_landmark) - ) - face_landmarks_results.append(face_landmarks_list) + face_landmarks = landmark_pb2.NormalizedLandmarkList() + face_landmarks.MergeFrom(face_landmarks_proto_list) + for face_landmark in face_landmarks.landmark: + holistic_landmarker_result.face_landmarks.append( + landmark_module.NormalizedLandmark.create_from_pb2(face_landmark) + ) - face_blendshapes_results = [] - if _FACE_BLENDSHAPES_STREAM_NAME in output_packets: - face_blendshapes_proto_list = packet_getter.get_proto_list( - output_packets[_FACE_BLENDSHAPES_STREAM_NAME] + pose_landmarks = landmark_pb2.NormalizedLandmarkList() + pose_landmarks.MergeFrom(pose_landmarks_proto_list) + for pose_landmark in pose_landmarks.landmark: + holistic_landmarker_result.pose_landmarks.append( + landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark) ) - for proto in face_blendshapes_proto_list: - face_blendshapes_categories = [] - face_blendshapes_classifications = classification_pb2.ClassificationList() - face_blendshapes_classifications.MergeFrom(proto) - for face_blendshapes in face_blendshapes_classifications.classification: - face_blendshapes_categories.append( - category_module.Category( - index=face_blendshapes.index, - score=face_blendshapes.score, - display_name=face_blendshapes.display_name, - category_name=face_blendshapes.label, - ) - ) - face_blendshapes_results.append(face_blendshapes_categories) - - for proto in pose_landmarks_proto_list: - pose_landmarks = landmark_pb2.NormalizedLandmarkList() - pose_landmarks.MergeFrom(proto) - pose_landmarks_list = [] - for pose_landmark in pose_landmarks.landmark: - pose_landmarks_list.append( - landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark) - ) - holistic_landmarker_result.pose_landmarks.append(pose_landmarks_list) - - for proto in pose_world_landmarks_proto_list: - pose_world_landmarks = landmark_pb2.LandmarkList() - pose_world_landmarks.MergeFrom(proto) - pose_world_landmarks_list = [] - for pose_world_landmark in pose_world_landmarks.landmark: - pose_world_landmarks_list.append( - landmark_module.Landmark.create_from_pb2(pose_world_landmark) - ) + + pose_world_landmarks = landmark_pb2.LandmarkList() + pose_world_landmarks.MergeFrom(pose_world_landmarks_proto_list) + for pose_world_landmark in pose_world_landmarks.landmark: holistic_landmarker_result.pose_world_landmarks.append( - pose_world_landmarks_list + landmark_module.Landmark.create_from_pb2(pose_world_landmark) ) - for proto in left_hand_landmarks_proto_list: - left_hand_landmarks = landmark_pb2.NormalizedLandmarkList() - left_hand_landmarks.MergeFrom(proto) - left_hand_landmarks_list = [] - for hand_landmark in left_hand_landmarks.landmark: - left_hand_landmarks_list.append( - landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) - ) + left_hand_landmarks = landmark_pb2.NormalizedLandmarkList() + left_hand_landmarks.MergeFrom(left_hand_landmarks_proto_list) + for hand_landmark in left_hand_landmarks.landmark: holistic_landmarker_result.left_hand_landmarks.append( - left_hand_landmarks_list + landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) ) - for proto in left_hand_world_landmarks_proto_list: - left_hand_world_landmarks = landmark_pb2.LandmarkList() - left_hand_world_landmarks.MergeFrom(proto) - left_hand_world_landmarks_list = [] - for left_hand_world_landmark in left_hand_world_landmarks.landmark: - left_hand_world_landmarks_list.append( - landmark_module.Landmark.create_from_pb2(left_hand_world_landmark) - ) + left_hand_world_landmarks = landmark_pb2.LandmarkList() + left_hand_world_landmarks.MergeFrom(left_hand_world_landmarks_proto_list) + for left_hand_world_landmark in left_hand_world_landmarks.landmark: holistic_landmarker_result.left_hand_world_landmarks.append( - left_hand_world_landmarks_list + landmark_module.Landmark.create_from_pb2(left_hand_world_landmark) ) - for proto in right_hand_landmarks_proto_list: - right_hand_landmarks = landmark_pb2.NormalizedLandmarkList() - right_hand_landmarks.MergeFrom(proto) - right_hand_landmarks_list = [] - for hand_landmark in right_hand_landmarks.landmark: - right_hand_landmarks_list.append( - landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) - ) + right_hand_landmarks = landmark_pb2.NormalizedLandmarkList() + right_hand_landmarks.MergeFrom(right_hand_landmarks_proto_list) + for hand_landmark in right_hand_landmarks.landmark: holistic_landmarker_result.right_hand_landmarks.append( - right_hand_landmarks_list + landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) ) - for proto in right_hand_world_landmarks_proto_list: - right_hand_world_landmarks = landmark_pb2.LandmarkList() - right_hand_world_landmarks.MergeFrom(proto) - right_hand_world_landmarks_list = [] - for right_hand_world_landmark in right_hand_world_landmarks.landmark: - right_hand_world_landmarks_list.append( - landmark_module.Landmark.create_from_pb2(right_hand_world_landmark) - ) + right_hand_world_landmarks = landmark_pb2.LandmarkList() + right_hand_world_landmarks.MergeFrom(right_hand_world_landmarks_proto_list) + for right_hand_world_landmark in right_hand_world_landmarks.landmark: holistic_landmarker_result.right_hand_world_landmarks.append( - right_hand_world_landmarks_list + landmark_module.Landmark.create_from_pb2(right_hand_world_landmark) + ) + + if _FACE_BLENDSHAPES_STREAM_NAME in output_packets: + face_blendshapes_proto_list = packet_getter.get_proto( + output_packets[_FACE_BLENDSHAPES_STREAM_NAME] + ) + face_blendshapes_classifications = classification_pb2.ClassificationList() + face_blendshapes_classifications.MergeFrom(face_blendshapes_proto_list) + holistic_landmarker_result.face_blendshapes = [] + for face_blendshapes in face_blendshapes_classifications.classification: + holistic_landmarker_result.face_blendshapes.append( + category_module.Category( + index=face_blendshapes.index, + score=face_blendshapes.score, + display_name=face_blendshapes.display_name, + category_name=face_blendshapes.label, + ) + ) + + if _POSE_SEGMENTATION_MASK_STREAM_NAME in output_packets: + holistic_landmarker_result.segmentation_masks = packet_getter.get_image_list( + output_packets[_POSE_SEGMENTATION_MASK_STREAM_NAME] ) return holistic_landmarker_result @@ -259,6 +271,9 @@ class HolisticLandmarkerOptions: landmark detection to be considered successful. min_hand_landmarks_confidence: The minimum confidence score for the hand landmark detection to be considered successful. + output_face_blendshapes: Whether FaceLandmarker outputs face blendshapes + classification. Face blendshapes are used for rendering the 3D face model. + output_segmentation_masks: whether to output segmentation masks. result_callback: The user-defined result callback for processing live stream data. The result callback should only be specified when the running mode is set to the live stream mode. @@ -419,7 +434,6 @@ def packets_callback(output_packets: Mapping[str, packet_module.Packet]): task_graph=_TASK_GRAPH_NAME, input_streams=[ ':'.join([_IMAGE_TAG, _IMAGE_IN_STREAM_NAME]), - ':'.join([_NORM_RECT_TAG, _NORM_RECT_STREAM_NAME]), ], output_streams=output_streams, task_options=options, @@ -436,7 +450,6 @@ def packets_callback(output_packets: Mapping[str, packet_module.Packet]): def detect( self, image: image_module.Image, - image_processing_options: Optional[_ImageProcessingOptions] = None, ) -> HolisticLandmarkerResult: """Performs holistic landmarks detection on the given image. @@ -449,7 +462,6 @@ def detect( Args: image: MediaPipe Image. - image_processing_options: Options for image processing. Returns: The holistic landmarks detection results. @@ -458,14 +470,8 @@ def detect( ValueError: If any of the input arguments is invalid. RuntimeError: If holistic landmarker detection failed to run. """ - normalized_rect = self.convert_to_normalized_rect( - image_processing_options, image, roi_allowed=False - ) output_packets = self._process_image_data({ _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image), - _NORM_RECT_STREAM_NAME: packet_creator.create_proto( - normalized_rect.to_pb2() - ), }) if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty(): @@ -477,7 +483,6 @@ def detect_for_video( self, image: image_module.Image, timestamp_ms: int, - image_processing_options: Optional[_ImageProcessingOptions] = None, ) -> HolisticLandmarkerResult: """Performs holistic landmarks detection on the provided video frame. @@ -492,7 +497,6 @@ def detect_for_video( Args: image: MediaPipe Image. timestamp_ms: The timestamp of the input video frame in milliseconds. - image_processing_options: Options for image processing. Returns: The holistic landmarks detection results. @@ -501,16 +505,10 @@ def detect_for_video( ValueError: If any of the input arguments is invalid. RuntimeError: If holistic landmarker detection failed to run. """ - normalized_rect = self.convert_to_normalized_rect( - image_processing_options, image, roi_allowed=False - ) output_packets = self._process_video_data({ _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at( timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND ), - _NORM_RECT_STREAM_NAME: packet_creator.create_proto( - normalized_rect.to_pb2() - ).at(timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND), }) if output_packets[_FACE_LANDMARKS_STREAM_NAME].is_empty(): @@ -522,7 +520,6 @@ def detect_async( self, image: image_module.Image, timestamp_ms: int, - image_processing_options: Optional[_ImageProcessingOptions] = None, ) -> None: """Sends live image data to perform holistic landmarks detection. @@ -548,20 +545,13 @@ def detect_async( Args: image: MediaPipe Image. timestamp_ms: The timestamp of the input image in milliseconds. - image_processing_options: Options for image processing. Raises: ValueError: If the current input timestamp is smaller than what the holistic landmarker has already processed. """ - normalized_rect = self.convert_to_normalized_rect( - image_processing_options, image, roi_allowed=False - ) self._send_live_stream_data({ _IMAGE_IN_STREAM_NAME: packet_creator.create_image(image).at( timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND ), - _NORM_RECT_STREAM_NAME: packet_creator.create_proto( - normalized_rect.to_pb2() - ).at(timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND), }) From 00eb1f190f2c2dd07879ba1fde179f14af24249d Mon Sep 17 00:00:00 2001 From: Kinar R <42828719+kinaryml@users.noreply.github.com> Date: Mon, 18 Dec 2023 04:57:25 +0530 Subject: [PATCH 4/9] Fix typo in holistic_landmarker_test.py --- mediapipe/tasks/python/test/vision/holistic_landmarker_test.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py index 6f27c903db..429ea98105 100644 --- a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py +++ b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py @@ -93,7 +93,7 @@ def setUp(self): def _expect_landmarks_correct( self, actual_landmarks, expected_landmarks, margin ): - # Expects to have the same number of poses detected. + # Expects to have the same number of landmarks detected. self.assertLen(actual_landmarks, len(expected_landmarks)) for i, elem in enumerate(actual_landmarks): From c0606e819c2920030585724f77e3c45ec8af277a Mon Sep 17 00:00:00 2001 From: Kinar R <42828719+kinaryml@users.noreply.github.com> Date: Mon, 18 Dec 2023 05:00:45 +0530 Subject: [PATCH 5/9] Update holistic_landmarker.py --- mediapipe/tasks/python/vision/holistic_landmarker.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py index 8edfaf445d..ac96ab7d15 100644 --- a/mediapipe/tasks/python/vision/holistic_landmarker.py +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -271,7 +271,7 @@ class HolisticLandmarkerOptions: landmark detection to be considered successful. min_hand_landmarks_confidence: The minimum confidence score for the hand landmark detection to be considered successful. - output_face_blendshapes: Whether FaceLandmarker outputs face blendshapes + output_face_blendshapes: Whether HolisticLandmarker outputs face blendshapes classification. Face blendshapes are used for rendering the 3D face model. output_segmentation_masks: whether to output segmentation masks. result_callback: The user-defined result callback for processing live stream From 30e6b766d47ac79a16f203dfbb22ea95ec0e0757 Mon Sep 17 00:00:00 2001 From: Kinar Date: Mon, 18 Dec 2023 02:47:28 -0800 Subject: [PATCH 6/9] Revised implementation and added more tests --- mediapipe/tasks/python/test/vision/BUILD | 5 - .../test/vision/holistic_landmarker_test.py | 386 ++++++++++++++++-- .../python/vision/holistic_landmarker.py | 140 ++++--- 3 files changed, 425 insertions(+), 106 deletions(-) diff --git a/mediapipe/tasks/python/test/vision/BUILD b/mediapipe/tasks/python/test/vision/BUILD index eae05de4d9..16a59741db 100644 --- a/mediapipe/tasks/python/test/vision/BUILD +++ b/mediapipe/tasks/python/test/vision/BUILD @@ -204,13 +204,8 @@ py_test( ], tags = ["not_run:arm"], deps = [ - "//mediapipe/framework/formats:classification_py_pb2", - "//mediapipe/framework/formats:landmark_py_pb2", "//mediapipe/tasks/cc/vision/holistic_landmarker/proto:holistic_result_py_pb2", "//mediapipe/python:_framework_bindings", - "//mediapipe/tasks/python/components/containers:category", - "//mediapipe/tasks/python/components/containers:landmark", - "//mediapipe/tasks/python/components/containers:rect", "//mediapipe/tasks/python/core:base_options", "//mediapipe/tasks/python/test:test_utils", "//mediapipe/tasks/python/vision:holistic_landmarker", diff --git a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py index 429ea98105..4b624af90f 100644 --- a/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py +++ b/mediapipe/tasks/python/test/vision/holistic_landmarker_test.py @@ -14,7 +14,6 @@ """Tests for holistic landmarker.""" import enum -from typing import List from unittest import mock from absl.testing import absltest @@ -22,13 +21,8 @@ import numpy as np from google.protobuf import text_format -from mediapipe.framework.formats import classification_pb2 -from mediapipe.framework.formats import landmark_pb2 from mediapipe.tasks.cc.vision.holistic_landmarker.proto import holistic_result_pb2 from mediapipe.python._framework_bindings import image as image_module -from mediapipe.tasks.python.components.containers import category as category_module -from mediapipe.tasks.python.components.containers import landmark as landmark_module -from mediapipe.tasks.python.components.containers import rect as rect_module from mediapipe.tasks.python.core import base_options as base_options_module from mediapipe.tasks.python.test import test_utils from mediapipe.tasks.python.vision import holistic_landmarker @@ -39,10 +33,6 @@ HolisticLandmarkerResult = holistic_landmarker.HolisticLandmarkerResult _HolisticResultProto = holistic_result_pb2.HolisticResult _BaseOptions = base_options_module.BaseOptions -_Category = category_module.Category -_Rect = rect_module.Rect -_Landmark = landmark_module.Landmark -_NormalizedLandmark = landmark_module.NormalizedLandmark _Image = image_module.Image _HolisticLandmarker = holistic_landmarker.HolisticLandmarker _HolisticLandmarkerOptions = holistic_landmarker.HolisticLandmarkerOptions @@ -53,23 +43,27 @@ _POSE_IMAGE = 'male_full_height_hands.jpg' _CAT_IMAGE = 'cat.jpg' _EXPECTED_HOLISTIC_RESULT = "male_full_height_hands_result_cpu.pbtxt" +_IMAGE_WIDTH = 638 +_IMAGE_HEIGHT = 1000 _LANDMARKS_MARGIN = 0.03 _BLENDSHAPES_MARGIN = 0.13 +_VIDEO_LANDMARKS_MARGIN = 0.03 +_VIDEO_BLENDSHAPES_MARGIN = 0.31 +_LIVE_STREAM_LANDMARKS_MARGIN = 0.03 +_LIVE_STREAM_BLENDSHAPES_MARGIN = 0.31 def _get_expected_holistic_landmarker_result( file_path: str, ) -> HolisticLandmarkerResult: - holistic_result_file_path = test_utils.get_test_data_path( - file_path - ) + holistic_result_file_path = test_utils.get_test_data_path(file_path) with open(holistic_result_file_path, 'rb') as f: holistic_result_proto = _HolisticResultProto() # Use this if a .pb file is available. # holistic_result_proto.ParseFromString(f.read()) text_format.Parse(f.read(), holistic_result_proto) holistic_landmarker_result = HolisticLandmarkerResult.create_from_pb2( - holistic_result_proto + holistic_result_proto ) return holistic_landmarker_result @@ -108,38 +102,70 @@ def _expect_blendshapes_correct( for i, elem in enumerate(actual_blendshapes): self.assertEqual(elem.index, expected_blendshapes[i].index) + self.assertEqual(elem.category_name, expected_blendshapes[i].category_name) self.assertAlmostEqual( - elem.score, - expected_blendshapes[i].score, - delta=margin, + elem.score, + expected_blendshapes[i].score, + delta=margin, ) def _expect_holistic_landmarker_results_correct( self, actual_result: HolisticLandmarkerResult, expected_result: HolisticLandmarkerResult, - output_segmentation_masks: bool, + output_segmentation_mask: bool, landmarks_margin: float, blendshapes_margin: float, ): self._expect_landmarks_correct( - actual_result.pose_landmarks, expected_result.pose_landmarks, - landmarks_margin + actual_result.pose_landmarks, expected_result.pose_landmarks, + landmarks_margin ) self._expect_landmarks_correct( - actual_result.face_landmarks, expected_result.face_landmarks, - landmarks_margin + actual_result.face_landmarks, expected_result.face_landmarks, + landmarks_margin ) self._expect_blendshapes_correct( - actual_result.face_blendshapes, expected_result.face_blendshapes, - blendshapes_margin + actual_result.face_blendshapes, expected_result.face_blendshapes, + blendshapes_margin ) - if output_segmentation_masks: - self.assertIsInstance(actual_result.segmentation_masks, List) - for _, mask in enumerate(actual_result.segmentation_masks): - self.assertIsInstance(mask, _Image) + if output_segmentation_mask: + self.assertIsInstance(actual_result.segmentation_mask, _Image) + self.assertEqual(actual_result.segmentation_mask.width, _IMAGE_WIDTH) + self.assertEqual(actual_result.segmentation_mask.height, _IMAGE_HEIGHT) else: - self.assertIsNone(actual_result.segmentation_masks) + self.assertIsNone(actual_result.segmentation_mask) + + def test_create_from_file_succeeds_with_valid_model_path(self): + # Creates with default option and valid model file successfully. + with _HolisticLandmarker.create_from_model_path(self.model_path) as landmarker: + self.assertIsInstance(landmarker, _HolisticLandmarker) + + def test_create_from_options_succeeds_with_valid_model_path(self): + # Creates with options containing model file successfully. + base_options = _BaseOptions(model_asset_path=self.model_path) + options = _HolisticLandmarkerOptions(base_options=base_options) + with _HolisticLandmarker.create_from_options(options) as landmarker: + self.assertIsInstance(landmarker, _HolisticLandmarker) + + def test_create_from_options_fails_with_invalid_model_path(self): + # Invalid empty model path. + with self.assertRaisesRegex( + RuntimeError, 'Unable to open file at /path/to/invalid/model.tflite' + ): + base_options = _BaseOptions( + model_asset_path='/path/to/invalid/model.tflite' + ) + options = _HolisticLandmarkerOptions(base_options=base_options) + _HolisticLandmarker.create_from_options(options) + + def test_create_from_options_succeeds_with_valid_model_content(self): + # Creates with options containing model content successfully. + with open(self.model_path, 'rb') as f: + base_options = _BaseOptions(model_asset_buffer=f.read()) + options = _HolisticLandmarkerOptions(base_options=base_options) + landmarker = _HolisticLandmarker.create_from_options(options) + self.assertIsInstance(landmarker, _HolisticLandmarker) @parameterized.parameters( ( @@ -154,13 +180,25 @@ def _expect_holistic_landmarker_results_correct( False, _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) ), + ( + ModelFileType.FILE_NAME, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + True, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ( + ModelFileType.FILE_CONTENT, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + True, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), ) def test_detect( self, model_file_type, model_name, - output_segmentation_masks, - expected_holistic_landmarker_result: HolisticLandmarkerResult + output_segmentation_mask, + expected_holistic_landmarker_result ): # Creates holistic landmarker. model_path = test_utils.get_test_data_path(model_name) @@ -178,7 +216,7 @@ def test_detect( base_options=base_options, output_face_blendshapes=True if expected_holistic_landmarker_result.face_blendshapes else False, - output_segmentation_masks=output_segmentation_masks, + output_segmentation_mask=output_segmentation_mask, ) landmarker = _HolisticLandmarker.create_from_options(options) @@ -186,12 +224,294 @@ def test_detect( detection_result = landmarker.detect(self.test_image) self._expect_holistic_landmarker_results_correct( detection_result, expected_holistic_landmarker_result, - output_segmentation_masks, _LANDMARKS_MARGIN, _BLENDSHAPES_MARGIN + output_segmentation_mask, _LANDMARKS_MARGIN, _BLENDSHAPES_MARGIN ) # Closes the holistic landmarker explicitly when the holistic landmarker is # not used in a context. landmarker.close() + @parameterized.parameters( + ( + ModelFileType.FILE_NAME, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + False, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ( + ModelFileType.FILE_CONTENT, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + True, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ) + def test_detect_in_context( + self, + model_file_type, + model_name, + output_segmentation_mask, + expected_holistic_landmarker_result + ): + # Creates holistic landmarker. + model_path = test_utils.get_test_data_path(model_name) + if model_file_type is ModelFileType.FILE_NAME: + base_options = _BaseOptions(model_asset_path=model_path) + elif model_file_type is ModelFileType.FILE_CONTENT: + with open(model_path, 'rb') as f: + model_content = f.read() + base_options = _BaseOptions(model_asset_buffer=model_content) + else: + # Should never happen + raise ValueError('model_file_type is invalid.') + + options = _HolisticLandmarkerOptions( + base_options=base_options, + output_face_blendshapes=True + if expected_holistic_landmarker_result.face_blendshapes else False, + output_segmentation_mask=output_segmentation_mask, + ) + + with _HolisticLandmarker.create_from_options(options) as landmarker: + # Performs holistic landmarks detection on the input. + detection_result = landmarker.detect(self.test_image) + self._expect_holistic_landmarker_results_correct( + detection_result, expected_holistic_landmarker_result, + output_segmentation_mask, _LANDMARKS_MARGIN, _BLENDSHAPES_MARGIN + ) + + def test_empty_detection_outputs(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path) + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + # Load the cat image. + cat_test_image = _Image.create_from_file( + test_utils.get_test_data_path(_CAT_IMAGE) + ) + # Performs holistic landmarks detection on the input. + detection_result = landmarker.detect(cat_test_image) + self.assertEmpty(detection_result.face_landmarks) + self.assertEmpty(detection_result.pose_landmarks) + self.assertEmpty(detection_result.pose_world_landmarks) + self.assertEmpty(detection_result.left_hand_landmarks) + self.assertEmpty(detection_result.left_hand_world_landmarks) + self.assertEmpty(detection_result.right_hand_landmarks) + self.assertEmpty(detection_result.right_hand_world_landmarks) + self.assertIsNone(detection_result.face_blendshapes) + self.assertIsNone(detection_result.segmentation_mask) + + def test_missing_result_callback(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.LIVE_STREAM, + ) + with self.assertRaisesRegex( + ValueError, r'result callback must be provided' + ): + with _HolisticLandmarker.create_from_options(options) as unused_landmarker: + pass + + @parameterized.parameters((_RUNNING_MODE.IMAGE), (_RUNNING_MODE.VIDEO)) + def test_illegal_result_callback(self, running_mode): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=running_mode, + result_callback=mock.MagicMock(), + ) + with self.assertRaisesRegex( + ValueError, r'result callback should not be provided' + ): + with _HolisticLandmarker.create_from_options(options) as unused_landmarker: + pass + + def test_calling_detect_for_video_in_image_mode(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.IMAGE, + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + with self.assertRaisesRegex( + ValueError, r'not initialized with the video mode' + ): + landmarker.detect_for_video(self.test_image, 0) + + def test_calling_detect_async_in_image_mode(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.IMAGE, + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + with self.assertRaisesRegex( + ValueError, r'not initialized with the live stream mode' + ): + landmarker.detect_async(self.test_image, 0) + + def test_calling_detect_in_video_mode(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.VIDEO, + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + with self.assertRaisesRegex( + ValueError, r'not initialized with the image mode' + ): + landmarker.detect(self.test_image) + + def test_calling_detect_async_in_video_mode(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.VIDEO, + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + with self.assertRaisesRegex( + ValueError, r'not initialized with the live stream mode' + ): + landmarker.detect_async(self.test_image, 0) + + def test_detect_for_video_with_out_of_order_timestamp(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.VIDEO, + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + unused_result = landmarker.detect_for_video(self.test_image, 1) + with self.assertRaisesRegex( + ValueError, r'Input timestamp must be monotonically increasing' + ): + landmarker.detect_for_video(self.test_image, 0) + + @parameterized.parameters( + ( + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + False, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ( + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + True, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ) + def test_detect_for_video( + self, + model_name, + output_segmentation_mask, + expected_holistic_landmarker_result + ): + # Creates holistic landmarker. + model_path = test_utils.get_test_data_path(model_name) + base_options = _BaseOptions(model_asset_path=model_path) + options = _HolisticLandmarkerOptions( + base_options=base_options, + running_mode=_RUNNING_MODE.VIDEO, + output_face_blendshapes=True + if expected_holistic_landmarker_result.face_blendshapes else False, + output_segmentation_mask=output_segmentation_mask, + ) + + with _HolisticLandmarker.create_from_options(options) as landmarker: + for timestamp in range(0, 300, 30): + # Performs holistic landmarks detection on the input. + detection_result = landmarker.detect_for_video( + self.test_image, timestamp + ) + # Comparing results. + self._expect_holistic_landmarker_results_correct( + detection_result, expected_holistic_landmarker_result, + output_segmentation_mask, + _VIDEO_LANDMARKS_MARGIN, _VIDEO_BLENDSHAPES_MARGIN + ) + + def test_calling_detect_in_live_stream_mode(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.LIVE_STREAM, + result_callback=mock.MagicMock(), + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + with self.assertRaisesRegex( + ValueError, r'not initialized with the image mode' + ): + landmarker.detect(self.test_image) + + def test_calling_detect_for_video_in_live_stream_mode(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.LIVE_STREAM, + result_callback=mock.MagicMock(), + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + with self.assertRaisesRegex( + ValueError, r'not initialized with the video mode' + ): + landmarker.detect_for_video(self.test_image, 0) + + def test_detect_async_calls_with_illegal_timestamp(self): + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=self.model_path), + running_mode=_RUNNING_MODE.LIVE_STREAM, + result_callback=mock.MagicMock(), + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + landmarker.detect_async(self.test_image, 100) + with self.assertRaisesRegex( + ValueError, r'Input timestamp must be monotonically increasing' + ): + landmarker.detect_async(self.test_image, 0) + + @parameterized.parameters( + ( + _POSE_IMAGE, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + False, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ( + _POSE_IMAGE, + _HOLISTIC_LANDMARKER_BUNDLE_ASSET_FILE, + True, + _get_expected_holistic_landmarker_result(_EXPECTED_HOLISTIC_RESULT) + ), + ) + def test_detect_async_calls( + self, + image_path, + model_name, + output_segmentation_mask, + expected_holistic_landmarker_result + ): + test_image = _Image.create_from_file( + test_utils.get_test_data_path(image_path) + ) + observed_timestamp_ms = -1 + + def check_result( + result: HolisticLandmarkerResult, output_image: _Image, timestamp_ms: int + ): + # Comparing results. + self._expect_holistic_landmarker_results_correct( + result, expected_holistic_landmarker_result, + output_segmentation_mask, + _LIVE_STREAM_LANDMARKS_MARGIN, _LIVE_STREAM_BLENDSHAPES_MARGIN + ) + self.assertTrue( + np.array_equal(output_image.numpy_view(), test_image.numpy_view()) + ) + self.assertLess(observed_timestamp_ms, timestamp_ms) + self.observed_timestamp_ms = timestamp_ms + + model_path = test_utils.get_test_data_path(model_name) + options = _HolisticLandmarkerOptions( + base_options=_BaseOptions(model_asset_path=model_path), + running_mode=_RUNNING_MODE.LIVE_STREAM, + output_face_blendshapes=True + if expected_holistic_landmarker_result.face_blendshapes else False, + output_segmentation_mask=output_segmentation_mask, + result_callback=check_result, + ) + with _HolisticLandmarker.create_from_options(options) as landmarker: + for timestamp in range(0, 300, 30): + landmarker.detect_async(test_image, timestamp) + if __name__ == '__main__': absltest.main() diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py index ac96ab7d15..315600b1c4 100644 --- a/mediapipe/tasks/python/vision/holistic_landmarker.py +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -51,7 +51,7 @@ _POSE_WORLD_LANDMARKS_STREAM_NAME = "pose_world_landmarks" _POSE_WORLD_LANDMARKS_TAG = "POSE_WORLD_LANDMARKS" _POSE_SEGMENTATION_MASK_STREAM_NAME = "pose_segmentation_mask" -_POSE_SEGMENTATION_MASK_TAG = "pose_segmentation_mask" +_POSE_SEGMENTATION_MASK_TAG = "POSE_SEGMENTATION_MASK" _FACE_LANDMARKS_STREAM_NAME = "face_landmarks" _FACE_LANDMARKS_TAG = "FACE_LANDMARKS" _FACE_BLENDSHAPES_STREAM_NAME = "extra_blendshapes" @@ -84,7 +84,7 @@ class HolisticLandmarkerResult: right_hand_landmarks: List[landmark_module.NormalizedLandmark] right_hand_world_landmarks: List[landmark_module.Landmark] face_blendshapes: Optional[List[category_module.Category]] = None - segmentation_masks: Optional[List[image_module.Image]] = None + segmentation_mask: Optional[image_module.Image] = None @classmethod @doc_controls.do_not_generate_docs @@ -96,41 +96,41 @@ def create_from_pb2( object.""" return HolisticLandmarkerResult( face_landmarks=[ - landmark_module.NormalizedLandmark.create_from_pb2(landmark) - for landmark in pb2_obj.face_landmarks.landmark + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.face_landmarks.landmark ] if hasattr(pb2_obj, 'face_landmarks') else None, pose_landmarks=[ - landmark_module.NormalizedLandmark.create_from_pb2(landmark) - for landmark in pb2_obj.pose_landmarks.landmark + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.pose_landmarks.landmark ] if hasattr(pb2_obj, 'pose_landmarks') else None, pose_world_landmarks=[ - landmark_module.Landmark.create_from_pb2(landmark) - for landmark in pb2_obj.pose_world_landmarks.landmark + landmark_module.Landmark.create_from_pb2(landmark) + for landmark in pb2_obj.pose_world_landmarks.landmark ] if hasattr(pb2_obj, 'pose_world_landmarks') else None, left_hand_landmarks=[ - landmark_module.NormalizedLandmark.create_from_pb2(landmark) - for landmark in pb2_obj.left_hand_landmarks.landmark + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.left_hand_landmarks.landmark ] if hasattr(pb2_obj, 'left_hand_landmarks') else None, left_hand_world_landmarks=[ - landmark_module.Landmark.create_from_pb2(landmark) - for landmark in pb2_obj.left_hand_world_landmarks.landmark + landmark_module.Landmark.create_from_pb2(landmark) + for landmark in pb2_obj.left_hand_world_landmarks.landmark ] if hasattr(pb2_obj, 'left_hand_world_landmarks') else None, right_hand_landmarks=[ - landmark_module.NormalizedLandmark.create_from_pb2(landmark) - for landmark in pb2_obj.right_hand_landmarks.landmark + landmark_module.NormalizedLandmark.create_from_pb2(landmark) + for landmark in pb2_obj.right_hand_landmarks.landmark ] if hasattr(pb2_obj, 'right_hand_landmarks') else None, right_hand_world_landmarks=[ - landmark_module.Landmark.create_from_pb2(landmark) - for landmark in pb2_obj.right_hand_world_landmarks.landmark + landmark_module.Landmark.create_from_pb2(landmark) + for landmark in pb2_obj.right_hand_world_landmarks.landmark ] if hasattr(pb2_obj, 'right_hand_world_landmarks') else None, face_blendshapes=[ - category_module.Category( - score=classification.score, - index=classification.index, - category_name=classification.label, - display_name=classification.display_name - ) - for classification in pb2_obj.face_blendshapes.classification + category_module.Category( + score=classification.score, + index=classification.index, + category_name=classification.label, + display_name=classification.display_name + ) + for classification in pb2_obj.face_blendshapes.classification ] if hasattr(pb2_obj, 'face_blendshapes') else None, ) @@ -147,98 +147,98 @@ def _build_landmarker_result( ) pose_landmarks_proto_list = packet_getter.get_proto( - output_packets[_POSE_LANDMARKS_STREAM_NAME] + output_packets[_POSE_LANDMARKS_STREAM_NAME] ) pose_world_landmarks_proto_list = packet_getter.get_proto( - output_packets[_POSE_WORLD_LANDMARKS_STREAM_NAME] + output_packets[_POSE_WORLD_LANDMARKS_STREAM_NAME] ) left_hand_landmarks_proto_list = packet_getter.get_proto( - output_packets[_LEFT_HAND_LANDMARKS_STREAM_NAME] + output_packets[_LEFT_HAND_LANDMARKS_STREAM_NAME] ) left_hand_world_landmarks_proto_list = packet_getter.get_proto( - output_packets[_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] + output_packets[_LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] ) right_hand_landmarks_proto_list = packet_getter.get_proto( - output_packets[_RIGHT_HAND_LANDMARKS_STREAM_NAME] + output_packets[_RIGHT_HAND_LANDMARKS_STREAM_NAME] ) right_hand_world_landmarks_proto_list = packet_getter.get_proto( - output_packets[_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] + output_packets[_RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] ) face_landmarks = landmark_pb2.NormalizedLandmarkList() face_landmarks.MergeFrom(face_landmarks_proto_list) for face_landmark in face_landmarks.landmark: holistic_landmarker_result.face_landmarks.append( - landmark_module.NormalizedLandmark.create_from_pb2(face_landmark) + landmark_module.NormalizedLandmark.create_from_pb2(face_landmark) ) pose_landmarks = landmark_pb2.NormalizedLandmarkList() pose_landmarks.MergeFrom(pose_landmarks_proto_list) for pose_landmark in pose_landmarks.landmark: holistic_landmarker_result.pose_landmarks.append( - landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark) + landmark_module.NormalizedLandmark.create_from_pb2(pose_landmark) ) pose_world_landmarks = landmark_pb2.LandmarkList() pose_world_landmarks.MergeFrom(pose_world_landmarks_proto_list) for pose_world_landmark in pose_world_landmarks.landmark: holistic_landmarker_result.pose_world_landmarks.append( - landmark_module.Landmark.create_from_pb2(pose_world_landmark) + landmark_module.Landmark.create_from_pb2(pose_world_landmark) ) left_hand_landmarks = landmark_pb2.NormalizedLandmarkList() left_hand_landmarks.MergeFrom(left_hand_landmarks_proto_list) for hand_landmark in left_hand_landmarks.landmark: holistic_landmarker_result.left_hand_landmarks.append( - landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) + landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) ) left_hand_world_landmarks = landmark_pb2.LandmarkList() left_hand_world_landmarks.MergeFrom(left_hand_world_landmarks_proto_list) for left_hand_world_landmark in left_hand_world_landmarks.landmark: holistic_landmarker_result.left_hand_world_landmarks.append( - landmark_module.Landmark.create_from_pb2(left_hand_world_landmark) + landmark_module.Landmark.create_from_pb2(left_hand_world_landmark) ) right_hand_landmarks = landmark_pb2.NormalizedLandmarkList() right_hand_landmarks.MergeFrom(right_hand_landmarks_proto_list) for hand_landmark in right_hand_landmarks.landmark: holistic_landmarker_result.right_hand_landmarks.append( - landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) + landmark_module.NormalizedLandmark.create_from_pb2(hand_landmark) ) right_hand_world_landmarks = landmark_pb2.LandmarkList() right_hand_world_landmarks.MergeFrom(right_hand_world_landmarks_proto_list) for right_hand_world_landmark in right_hand_world_landmarks.landmark: holistic_landmarker_result.right_hand_world_landmarks.append( - landmark_module.Landmark.create_from_pb2(right_hand_world_landmark) + landmark_module.Landmark.create_from_pb2(right_hand_world_landmark) ) if _FACE_BLENDSHAPES_STREAM_NAME in output_packets: face_blendshapes_proto_list = packet_getter.get_proto( - output_packets[_FACE_BLENDSHAPES_STREAM_NAME] + output_packets[_FACE_BLENDSHAPES_STREAM_NAME] ) face_blendshapes_classifications = classification_pb2.ClassificationList() face_blendshapes_classifications.MergeFrom(face_blendshapes_proto_list) holistic_landmarker_result.face_blendshapes = [] for face_blendshapes in face_blendshapes_classifications.classification: holistic_landmarker_result.face_blendshapes.append( - category_module.Category( - index=face_blendshapes.index, - score=face_blendshapes.score, - display_name=face_blendshapes.display_name, - category_name=face_blendshapes.label, - ) + category_module.Category( + index=face_blendshapes.index, + score=face_blendshapes.score, + display_name=face_blendshapes.display_name, + category_name=face_blendshapes.label, + ) ) if _POSE_SEGMENTATION_MASK_STREAM_NAME in output_packets: - holistic_landmarker_result.segmentation_masks = packet_getter.get_image_list( - output_packets[_POSE_SEGMENTATION_MASK_STREAM_NAME] + holistic_landmarker_result.segmentation_mask = packet_getter.get_image( + output_packets[_POSE_SEGMENTATION_MASK_STREAM_NAME] ) return holistic_landmarker_result @@ -273,7 +273,7 @@ class HolisticLandmarkerOptions: landmark detection to be considered successful. output_face_blendshapes: Whether HolisticLandmarker outputs face blendshapes classification. Face blendshapes are used for rendering the 3D face model. - output_segmentation_masks: whether to output segmentation masks. + output_segmentation_mask: whether to output segmentation masks. result_callback: The user-defined result callback for processing live stream data. The result callback should only be specified when the running mode is set to the live stream mode. @@ -290,7 +290,7 @@ class HolisticLandmarkerOptions: min_pose_landmarks_confidence: float = 0.5 min_hand_landmarks_confidence: float = 0.5 output_face_blendshapes: bool = False - output_segmentation_masks: bool = False + output_segmentation_mask: bool = False result_callback: Optional[ Callable[[HolisticLandmarkerResult, image_module.Image, int], None] ] = None @@ -319,17 +319,17 @@ def to_pb2(self) -> _HolisticLandmarkerGraphOptionsProto: ) # Configure pose detector and pose landmarks detector options. holistic_landmarker_options_proto.pose_detector_graph_options.min_detection_confidence = ( - self.min_pose_detection_confidence + self.min_pose_detection_confidence ) holistic_landmarker_options_proto.pose_detector_graph_options.min_suppression_threshold = ( - self.min_pose_suppression_threshold + self.min_pose_suppression_threshold ) holistic_landmarker_options_proto.face_landmarks_detector_graph_options.min_detection_confidence = ( - self.min_pose_landmarks_confidence + self.min_pose_landmarks_confidence ) # Configure hand landmarks detector options. holistic_landmarker_options_proto.hand_landmarks_detector_graph_options.min_detection_confidence = ( - self.min_hand_landmarks_confidence + self.min_hand_landmarks_confidence ) return holistic_landmarker_options_proto @@ -404,30 +404,34 @@ def packets_callback(output_packets: Mapping[str, packet_module.Packet]): ) output_streams = [ - ':'.join([_FACE_LANDMARKS_TAG, _FACE_LANDMARKS_STREAM_NAME]), - ':'.join([_POSE_LANDMARKS_TAG_NAME, _POSE_LANDMARKS_STREAM_NAME]), - ':'.join( - [_POSE_WORLD_LANDMARKS_TAG, _POSE_WORLD_LANDMARKS_STREAM_NAME] - ), - ':'.join([_LEFT_HAND_LANDMARKS_TAG, _LEFT_HAND_LANDMARKS_STREAM_NAME]), - ':'.join( - [_LEFT_HAND_WORLD_LANDMARKS_TAG, _LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] - ), - ':'.join([_RIGHT_HAND_LANDMARKS_TAG, _RIGHT_HAND_LANDMARKS_STREAM_NAME]), - ':'.join( - [_RIGHT_HAND_WORLD_LANDMARKS_TAG, _RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] - ), - ':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME]), + ':'.join([_FACE_LANDMARKS_TAG, _FACE_LANDMARKS_STREAM_NAME]), + ':'.join([_POSE_LANDMARKS_TAG_NAME, _POSE_LANDMARKS_STREAM_NAME]), + ':'.join( + [_POSE_WORLD_LANDMARKS_TAG, _POSE_WORLD_LANDMARKS_STREAM_NAME] + ), + ':'.join([_LEFT_HAND_LANDMARKS_TAG, _LEFT_HAND_LANDMARKS_STREAM_NAME]), + ':'.join( + [_LEFT_HAND_WORLD_LANDMARKS_TAG, + _LEFT_HAND_WORLD_LANDMARKS_STREAM_NAME] + ), + ':'.join([_RIGHT_HAND_LANDMARKS_TAG, + _RIGHT_HAND_LANDMARKS_STREAM_NAME]), + ':'.join( + [_RIGHT_HAND_WORLD_LANDMARKS_TAG, + _RIGHT_HAND_WORLD_LANDMARKS_STREAM_NAME] + ), + ':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME]), ] - if options.output_segmentation_masks: + if options.output_segmentation_mask: output_streams.append( - ':'.join([_POSE_SEGMENTATION_MASK_TAG, _POSE_SEGMENTATION_MASK_STREAM_NAME]) + ':'.join([_POSE_SEGMENTATION_MASK_TAG, + _POSE_SEGMENTATION_MASK_STREAM_NAME]) ) if options.output_face_blendshapes: output_streams.append( - ':'.join([_FACE_BLENDSHAPES_TAG, _FACE_BLENDSHAPES_STREAM_NAME]) + ':'.join([_FACE_BLENDSHAPES_TAG, _FACE_BLENDSHAPES_STREAM_NAME]) ) task_info = _TaskInfo( From a667949b99772f074d2bf8b99decd6b80d05dab7 Mon Sep 17 00:00:00 2001 From: Kinar Date: Mon, 18 Dec 2023 02:49:20 -0800 Subject: [PATCH 7/9] Changed import style in task_info --- mediapipe/tasks/python/core/task_info.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/mediapipe/tasks/python/core/task_info.py b/mediapipe/tasks/python/core/task_info.py index 5d039a0341..96bb5dac93 100644 --- a/mediapipe/tasks/python/core/task_info.py +++ b/mediapipe/tasks/python/core/task_info.py @@ -21,7 +21,7 @@ from mediapipe.framework import calculator_options_pb2 from mediapipe.framework import calculator_pb2 from mediapipe.tasks.python.core.optional_dependencies import doc_controls -from google.protobuf.any_pb2 import Any +from google.protobuf import any_pb2 @doc_controls.do_not_generate_docs @@ -98,7 +98,7 @@ def add_stream_name_prefix(tag_index_name): node_config.options.CopyFrom(task_subgraph_options) else: # Use the Any type for task_subgraph_options (proto3) - task_subgraph_options = Any() + task_subgraph_options = any_pb2.Any() task_subgraph_options.Pack(self.task_options.to_pb2()) node_config.node_options.append(task_subgraph_options) From cba08786529822933352e7829779fe3a033801e7 Mon Sep 17 00:00:00 2001 From: Kinar Date: Mon, 18 Dec 2023 02:51:11 -0800 Subject: [PATCH 8/9] Updated Copyright year --- mediapipe/tasks/python/vision/holistic_landmarker.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py index 315600b1c4..a7a6499748 100644 --- a/mediapipe/tasks/python/vision/holistic_landmarker.py +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -1,4 +1,4 @@ -# Copyright 2022 The MediaPipe Authors. +# Copyright 2023 The MediaPipe Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. From 569c16db0b07de2666f39f8c49b32127c804d0f6 Mon Sep 17 00:00:00 2001 From: Kinar Date: Mon, 18 Dec 2023 04:37:29 -0800 Subject: [PATCH 9/9] Documented HolisticLandmarkerResult --- .../tasks/python/vision/holistic_landmarker.py | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) diff --git a/mediapipe/tasks/python/vision/holistic_landmarker.py b/mediapipe/tasks/python/vision/holistic_landmarker.py index a7a6499748..27433c1d66 100644 --- a/mediapipe/tasks/python/vision/holistic_landmarker.py +++ b/mediapipe/tasks/python/vision/holistic_landmarker.py @@ -74,7 +74,19 @@ class HolisticLandmarkerResult: """The holistic landmarks result from HolisticLandmarker, where each vector element represents a single holistic detected in the image. Attributes: - TODO + face_landmarks: Detected face landmarks in normalized image coordinates. + pose_landmarks: Detected pose landmarks in normalized image coordinates. + pose_world_landmarks: Detected pose world landmarks in image coordinates. + left_hand_landmarks: Detected left hand landmarks in normalized image + coordinates. + left_hand_world_landmarks: Detected left hand landmarks in image + coordinates. + right_hand_landmarks: Detected right hand landmarks in normalized image + coordinates. + right_hand_world_landmarks: Detected right hand landmarks in image + coordinates. + face_blendshapes: Optional face blendshapes. + segmentation_mask: Optional segmentation mask for pose. """ face_landmarks: List[landmark_module.NormalizedLandmark] pose_landmarks: List[landmark_module.NormalizedLandmark]