Move library into own folder

.github/workflows/python.yml

@@ -25,7 +25,7 @@ jobs:
       run: |
         python --version
         echo "import tests.actions_test" > test.py
-        coverage run --source models,examples test.py
+        coverage run --source facenet_pytorch/models,examples tests/actions_test.py
         coverage report
         coverage xml
     - name: Upload coverage to Codecov

README.md

@@ -1,7 +1,9 @@
-# Face Recognition Using Pytorch 
+# Face Recognition Using Pytorch
 
 *You can also read a translated version of this file [in Chinese 简体中文版](README_cn.md).*
 
+![banner](static/facenet-pytorch-banner.png)
+
 [![Downloads](https://pepy.tech/badge/facenet-pytorch)](https://pepy.tech/project/facenet-pytorch)
 
 [![Code Coverage](https://img.shields.io/codecov/c/github/timesler/facenet-pytorch.svg)](https://codecov.io/gh/timesler/facenet-pytorch)
@@ -18,12 +20,12 @@ Also included in this repo is an efficient pytorch implementation of MTCNN for f
 * [Quick start](#quick-start)
 * [Pretrained models](#pretrained-models)
 * [Example notebooks](#example-notebooks)
-  + [*Complete detection and recognition pipeline*](#complete-detection-and-recognition-pipeline)
-  + [*Face tracking in video streams*](#face-tracking-in-video-streams)
-  + [*Finetuning pretrained models with new data*](#finetuning-pretrained-models-with-new-data)
-  + [*Guide to MTCNN in facenet-pytorch*](#guide-to-mtcnn-in-facenet-pytorch)
-  + [*Performance comparison of face detection packages*](#performance-comparison-of-face-detection-packages)
-  + [*The FastMTCNN algorithm*](#the-fastmtcnn-algorithm)
+  * [*Complete detection and recognition pipeline*](#complete-detection-and-recognition-pipeline)
+  * [*Face tracking in video streams*](#face-tracking-in-video-streams)
+  * [*Finetuning pretrained models with new data*](#finetuning-pretrained-models-with-new-data)
+  * [*Guide to MTCNN in facenet-pytorch*](#guide-to-mtcnn-in-facenet-pytorch)
+  * [*Performance comparison of face detection packages*](#performance-comparison-of-face-detection-packages)
+  * [*The FastMTCNN algorithm*](#the-fastmtcnn-algorithm)
 * [Running with docker](#running-with-docker)
 * [Use this repo in your own git project](#use-this-repo-in-your-own-git-project)
 * [Conversion of parameters from Tensorflow to Pytorch](#conversion-of-parameters-from-tensorflow-to-pytorch)
@@ -32,7 +34,7 @@ Also included in this repo is an efficient pytorch implementation of MTCNN for f
 ## Quick start
 
 1. Install:
-    
+
     ```bash
     # With pip:
     pip install facenet-pytorch
@@ -43,10 +45,10 @@ Also included in this repo is an efficient pytorch implementation of MTCNN for f
     # or use a docker container (see https://github.com/timesler/docker-jupyter-dl-gpu):
     docker run -it --rm timesler/jupyter-dl-gpu pip install facenet-pytorch && ipython
     ```
-    
+
 1. In python, import facenet-pytorch and instantiate models:
-    
-    ```python
+
+    ```py
     from facenet_pytorch import MTCNN, InceptionResnetV1
     
     # If required, create a face detection pipeline using MTCNN:
@@ -55,10 +57,10 @@ Also included in this repo is an efficient pytorch implementation of MTCNN for f
     # Create an inception resnet (in eval mode):
     resnet = InceptionResnetV1(pretrained='vggface2').eval()
     ```
-    
+
 1. Process an image:
-    
-    ```python
+
+    ```py
     from PIL import Image
     
     img = Image.open(<image path>)
@@ -78,7 +80,7 @@ See `help(MTCNN)` and `help(InceptionResnetV1)` for usage and implementation det
 
 ## Pretrained models
 
-See: [models/inception_resnet_v1.py](models/inception_resnet_v1.py)
+See: [models/inception_resnet_v1.py](facenet_pytorch/models/inception_resnet_v1.py)
 
 The following models have been ported to pytorch (with links to download pytorch state_dict's):
 
@@ -89,7 +91,7 @@ The following models have been ported to pytorch (with links to download pytorch
 
 There is no need to manually download the pretrained state_dict's; they are downloaded automatically on model instantiation and cached for future use in the torch cache. To use an Inception Resnet (V1) model for facial recognition/identification in pytorch, use:
 
-```python
+```py
 from facenet_pytorch import InceptionResnetV1
 
 # For a model pretrained on VGGFace2
@@ -119,7 +121,7 @@ Face recognition can be easily applied to raw images by first detecting faces us
 
 MTCNN can be used to build a face tracking system (using the `MTCNN.detect()` method). A full face tracking example can be found at [examples/face_tracking.ipynb](examples/face_tracking.ipynb).
 
-![](examples/tracked.gif)
+![Tracking](examples/tracked.gif)
 
 ### *Finetuning pretrained models with new data*
 
@@ -156,7 +158,7 @@ Each package is tested for its speed in detecting the faces in a set of 300 imag
 |dlib|3.80|8.39|14.53|
 |mtcnn|3.04|5.70|8.23|
 
-![](examples/performance-comparison.png)
+![Performance comparison](examples/performance-comparison.png)
 
 See the [notebook on kaggle](https://www.kaggle.com/timesler/comparison-of-face-detection-packages).
 
@@ -193,9 +195,9 @@ Alternatively, the code can be installed as a package using pip:
 
 ## Conversion of parameters from Tensorflow to Pytorch
 
-See: [models/utils/tensorflow2pytorch.py](models/tensorflow2pytorch.py)
+See: [models/utils/tensorflow2pytorch.py](facenet_pytorch/models/tensorflow2pytorch.py)
 
-Note that this functionality is not needed to use the models in this repo, which depend only on the saved pytorch `state_dict`'s. 
+Note that this functionality is not needed to use the models in this repo, which depend only on the saved pytorch `state_dict`'s.
 
 Following instantiation of the pytorch model, each layer's weights were loaded from equivalent layers in the pretrained tensorflow models from [davidsandberg/facenet](https://github.com/davidsandberg/facenet).
 
@@ -205,7 +207,7 @@ The equivalence of the outputs from the original tensorflow models and the pytor
 
 `>>> compare_model_outputs(mdl, sess, torch.randn(5, 160, 160, 3).detach())`
 
-```
+```py
 Passing test data through TF model
 
 tensor([[-0.0142,  0.0615,  0.0057,  ...,  0.0497,  0.0375, -0.0838],
@@ -228,16 +230,16 @@ Distance 1.2874517096861382e-06
 
 ---
 
-In order to re-run the conversion of tensorflow parameters into the pytorch model, ensure you clone this repo _with submodules_, as the davidsandberg/facenet repo is included as a submodule and parts of it are required for the conversion.
+In order to re-run the conversion of tensorflow parameters into the pytorch model, ensure you clone this repo *with submodules_, as the davidsandberg/facenet repo is included as a submodule and parts of it are required for the conversion.
 
 ## References
 
 1. David Sandberg's facenet repo: [https://github.com/davidsandberg/facenet](https://github.com/davidsandberg/facenet)
 
-1. F. Schroff, D. Kalenichenko, J. Philbin. _FaceNet: A Unified Embedding for Face Recognition and Clustering_, arXiv:1503.03832, 2015. [PDF](https://arxiv.org/pdf/1503.03832)
+1. F. Schroff, D. Kalenichenko, J. Philbin. *FaceNet: A Unified Embedding for Face Recognition and Clustering_, arXiv:1503.03832, 2015. [PDF](https://arxiv.org/pdf/1503.03832)
 
-1. Q. Cao, L. Shen, W. Xie, O. M. Parkhi, A. Zisserman. _VGGFace2: A dataset for recognising face across pose and age_, International Conference on Automatic Face and Gesture Recognition, 2018. [PDF](http://www.robots.ox.ac.uk/~vgg/publications/2018/Cao18/cao18.pdf)
+1. Q. Cao, L. Shen, W. Xie, O. M. Parkhi, A. Zisserman. *VGGFace2: A dataset for recognising face across pose and age_, International Conference on Automatic Face and Gesture Recognition, 2018. [PDF](http://www.robots.ox.ac.uk/~vgg/publications/2018/Cao18/cao18.pdf)
 
-1. D. Yi, Z. Lei, S. Liao and S. Z. Li. _CASIAWebface: Learning Face Representation from Scratch_, arXiv:1411.7923, 2014. [PDF](https://arxiv.org/pdf/1411.7923)
+1. D. Yi, Z. Lei, S. Liao and S. Z. Li. *CASIAWebface: Learning Face Representation from Scratch_, arXiv:1411.7923, 2014. [PDF](https://arxiv.org/pdf/1411.7923)
 
-1. K. Zhang, Z. Zhang, Z. Li and Y. Qiao. _Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks_, IEEE Signal Processing Letters, 2016. [PDF](https://kpzhang93.github.io/MTCNN_face_detection_alignment/paper/spl.pdf)
+1. K. Zhang, Z. Zhang, Z. Li and Y. Qiao. *Joint Face Detection and Alignment Using Multitask Cascaded Convolutional Networks_, IEEE Signal Processing Letters, 2016. [PDF](https://kpzhang93.github.io/MTCNN_face_detection_alignment/paper/spl.pdf)

README_cn.md

@@ -118,7 +118,7 @@ img_probs = resnet(img_cropped.unsqueeze(0))
 
 ## 预训练模型
 
-请参阅：[models/inception_resnet_v1.py](models/inception_resnet_v1.py)
+请参阅：[models/inception_resnet_v1.py](facenet_pytorch/models/inception_resnet_v1.py)
 
 以下模型已移植到 pytorch（包含下载 pytorch state_dict 的链接）：
 
@@ -233,7 +233,7 @@ docker run --rm -p 8888:8888
 
 ## Tensorflow 到 Pytorch 的参数转换
 
-请参阅：[models/utils/tensorflow2pytorch.py](models/tensorflow2pytorch.py)
+请参阅：[models/utils/tensorflow2pytorch.py](facenet_pytorch/models/tensorflow2pytorch.py)
 
 请注意，使用此存储库中的模型不需要此功能，该功能仅依赖于pytorch保存的 `state_dict`。
 

examples/.gitignore

@@ -0,0 +1 @@
+tmptest.py

examples/finetune.ipynb

@@ -15,6 +15,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from pathlib import Path\n",
     "from facenet_pytorch import MTCNN, InceptionResnetV1, fixed_image_standardization, training\n",
     "import torch\n",
     "from torch.utils.data import DataLoader, SubsetRandomSampler\n",
@@ -41,7 +42,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "data_dir = '../data/test_images'\n",
+    "root_dir = Path(__file__).parent.parent\n",
+    "data_dir = root_dir / \"tests\" / \"data\" / \"test_images\"\n",
     "\n",
     "batch_size = 32\n",
     "epochs = 8\n",
@@ -106,19 +108,19 @@
    "source": [
     "dataset = datasets.ImageFolder(data_dir, transform=transforms.Resize((512, 512)))\n",
     "dataset.samples = [\n",
-    "    (p, p.replace(data_dir, data_dir + '_cropped'))\n",
+    "    (p, p.replace(str(data_dir), str(data_dir.parent / (data_dir.name + \"_aligned\"))))\n",
     "        for p, _ in dataset.samples\n",
     "]\n",
-    "        \n",
+    "\n",
     "loader = DataLoader(\n",
     "    dataset,\n",
     "    num_workers=workers,\n",
     "    batch_size=batch_size,\n",
     "    collate_fn=training.collate_pil\n",
     ")\n",
     "\n",
-    "for i, (x, y) in enumerate(loader):\n",
-    "    mtcnn(x, save_path=y)\n",
+    "for i, (input_file, output_file) in enumerate(loader):\n",
+    "    mtcnn(input_file, save_path=output_file)\n",
     "    print('\\rBatch {} of {}'.format(i + 1, len(loader)), end='')\n",
     "    \n",
     "# Remove mtcnn to reduce GPU memory usage\n",
@@ -168,7 +170,7 @@
     "    transforms.ToTensor(),\n",
     "    fixed_image_standardization\n",
     "])\n",
-    "dataset = datasets.ImageFolder(data_dir + '_cropped', transform=trans)\n",
+    "dataset = datasets.ImageFolder(data_dir.parent / (data_dir.name + \"_aligned\"), transform=trans)\n",
     "img_inds = np.arange(len(dataset))\n",
     "np.random.shuffle(img_inds)\n",
     "train_inds = img_inds[:int(0.8 * len(img_inds))]\n",

examples/finetune_cn.ipynb

@@ -15,6 +15,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from pathlib import Path\n",
     "from facenet_pytorch import MTCNN, InceptionResnetV1, fixed_image_standardization, training\n",
     "import torch\n",
     "from torch.utils.data import DataLoader, SubsetRandomSampler\n",
@@ -41,7 +42,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "data_dir = '../data/test_images'\n",
+    "root_dir = Path(__file__).parent.parent\n",
+    "data_dir = root_dir / \"tests\" / \"data\" / \"test_images\"\n",
     "\n",
     "batch_size = 32\n",
     "epochs = 8\n",
@@ -106,7 +108,7 @@
    "source": [
     "dataset = datasets.ImageFolder(data_dir, transform=transforms.Resize((512, 512)))\n",
     "dataset.samples = [\n",
-    "    (p, p.replace(data_dir, data_dir + '_cropped'))\n",
+    "    (p, p.replace(str(data_dir), str(data_dir.parent / (data_dir.name + \"_aligned\"))))\n",
     "        for p, _ in dataset.samples\n",
     "]\n",
     "        \n",
@@ -168,7 +170,7 @@
     "    transforms.ToTensor(),\n",
     "    fixed_image_standardization\n",
     "])\n",
-    "dataset = datasets.ImageFolder(data_dir + '_cropped', transform=trans)\n",
+    "dataset = datasets.ImageFolder(data_dir + '_aligned', transform=trans)\n",
     "img_inds = np.arange(len(dataset))\n",
     "np.random.shuffle(img_inds)\n",
     "train_inds = img_inds[:int(0.8 * len(img_inds))]\n",

examples/infer.ipynb

@@ -20,13 +20,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
+    "from pathlib import Path\n",
+    "import os\n",
+    "\n",
     "from facenet_pytorch import MTCNN, InceptionResnetV1\n",
     "import torch\n",
     "from torch.utils.data import DataLoader\n",
     "from torchvision import datasets\n",
-    "import numpy as np\n",
     "import pandas as pd\n",
-    "import os\n",
     "\n",
     "workers = 0 if os.name == 'nt' else 4"
    ]
@@ -118,7 +119,10 @@
     "def collate_fn(x):\n",
     "    return x[0]\n",
     "\n",
-    "dataset = datasets.ImageFolder('../data/test_images')\n",
+    "root_dir = Path(__file__).parent.parent\n",
+    "data_dir = root_dir / \"tests\" / \"data\" / \"test_images\"\n",
+    "\n",
+    "dataset = datasets.ImageFolder(data_dir)\n",
     "dataset.idx_to_class = {i:c for c, i in dataset.class_to_idx.items()}\n",
     "loader = DataLoader(dataset, collate_fn=collate_fn, num_workers=workers)"
    ]
@@ -242,4 +246,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 2
-}
+}

examples/infer_cn.ipynb

@@ -26,13 +26,14 @@
    },
    "outputs": [],
    "source": [
+    "from pathlib import Path\n",
+    "import os\n",
+    "\n",
     "from facenet_pytorch import MTCNN, InceptionResnetV1\n",
     "import torch\n",
     "from torch.utils.data import DataLoader\n",
     "from torchvision import datasets\n",
-    "import numpy as np\n",
     "import pandas as pd\n",
-    "import os\n",
     "\n",
     "workers = 0 if os.name == 'nt' else 4"
    ]
@@ -144,7 +145,10 @@
     "def collate_fn(x):\n",
     "    return x[0]\n",
     "\n",
-    "dataset = datasets.ImageFolder('../data/test_images')\n",
+    "root_dir = Path(__file__).parent.parent\n",
+    "data_dir = root_dir / \"tests\" / \"data\" / \"test_images\"\n",
+    "\n",
+    "dataset = datasets.ImageFolder(data_dir)\n",
     "dataset.idx_to_class = {i:c for c, i in dataset.class_to_idx.items()}\n",
     "loader = DataLoader(dataset, collate_fn=collate_fn, num_workers=workers)"
    ]

examples/lfw_evaluate.ipynb

@@ -141,7 +141,7 @@
     "box_probs = []\n",
     "\n",
     "for i, (x, b_paths) in enumerate(loader):\n",
-    "    crops = [p.replace(data_dir, data_dir + '_cropped') for p in b_paths]\n",
+    "    crops = [p.replace(str(data_dir), str(data_dir.parent / (data_dir.name + \"_aligned\"))) for p in b_paths]\n",
     "    mtcnn(x, save_path=crops)\n",
     "    crop_paths.extend(crops)\n",
     "    print('\\rBatch {} of {}'.format(i + 1, len(loader)), end='')"
@@ -176,7 +176,7 @@
     "    fixed_image_standardization\n",
     "])\n",
     "\n",
-    "dataset = datasets.ImageFolder(data_dir + '_cropped', transform=trans)\n",
+    "dataset = datasets.ImageFolder(data_dir.parent / data_dir.name + \"_aligned\", transform=trans)\n",
     "\n",
     "embed_loader = DataLoader(\n",
     "    dataset,\n",
@@ -452,7 +452,7 @@
    "outputs": [],
    "source": [
     "pairs = read_pairs(pairs_path)\n",
-    "path_list, issame_list = get_paths(data_dir+'_cropped', pairs)\n",
+    "path_list, issame_list = get_paths(data_dir+'_aligned', pairs)\n",
     "embeddings = np.array([embeddings_dict[path] for path in path_list])\n",
     "\n",
     "tpr, fpr, accuracy, val, val_std, far, fp, fn = evaluate(embeddings, issame_list)"
@@ -519,4 +519,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 0
-}
+}