fix issue #I8B8SV, #I7RZSB, and bug of exporting visual results

deploy/py_infer/src/utils/visual_utils.py

@@ -69,14 +69,15 @@ def vis_bbox_text(image, box_list, text_list, font_path):
 def draw_box_txt_fine(img_size, box, txt, font_path):
     box_height = int(math.sqrt((box[0][0] - box[3][0]) ** 2 + (box[0][1] - box[3][1]) ** 2))
     box_width = int(math.sqrt((box[0][0] - box[1][0]) ** 2 + (box[0][1] - box[1][1]) ** 2))
-    img_text = Image.new("RGB", (box_width, box_height), (255, 255, 255))  # RGB or BGR doesn't matter
     if box_height > 2 * box_width and box_height > 30:
+        img_text = Image.new("RGB", (box_height, box_width), (255, 255, 255))  # RGB or BGR doesn't matter
         draw_text = ImageDraw.Draw(img_text)
         if txt:
             font = create_font(txt, (box_height, box_width), font_path)
             draw_text.text([0, 0], txt, fill=(0, 0, 0), font=font)
         img_text = img_text.transpose(Image.ROTATE_270)
     else:
+        img_text = Image.new("RGB", (box_width, box_height), (255, 255, 255))  # RGB or BGR doesn't matter
         draw_text = ImageDraw.Draw(img_text)
         if txt:
             font = create_font(txt, (box_width, box_height), font_path)

docs/cn/datasets/ccpd.md

@@ -35,8 +35,8 @@ CCPD2019
 ```bash
 python tools/dataset_converters/convert.py \
     --dataset_name ccpd --task det \
-    --image_dir path/to/CCPD2019/ \
-    --label_dir path/to/CCPD2019/splits/train.txt
+    --image_dir path/to/CCPD2019 \
+    --label_dir path/to/CCPD2019/splits/train.txt \
     --output_path path/to/CCPD2019/det_gt.txt
 ```
 

docs/en/datasets/ccpd.md

@@ -34,7 +34,7 @@ To prepare the data for text detection, you can run the following commands:
 ```bash
 python tools/dataset_converters/convert.py \
     --dataset_name ccpd --task det \
-    --image_dir path/to/CCPD2019/ \
+    --image_dir path/to/CCPD2019 \
     --label_dir path/to/CCPD2019/splits/train.txt \
     --output_path path/to/CCPD2019/det_gt.txt
 ```

mindocr/optim/adamw.py

@@ -3,7 +3,7 @@
 import numpy as np
 
 import mindspore as ms
-from mindspore import ops
+from mindspore import ops, version
 from mindspore.common.initializer import initializer
 from mindspore.common.parameter import Parameter
 from mindspore.common.tensor import Tensor
@@ -174,6 +174,8 @@ def __init__(
     @jit
     def construct(self, gradients):
         lr = self.get_lr()
+        if version.__version__ >= "2.2" and self._is_dynamic_lr_or_weight_decay():
+            self.assignadd(self.global_step, self.global_step_increase_tensor)
         gradients = scale_grad(gradients, self.reciprocal_scale)
         if self.clip:
             gradients = ops.clip_by_global_norm(gradients, 5.0, None)

mindocr/optim/adan.py

@@ -1,6 +1,6 @@
 """adan"""
 import mindspore as ms
-from mindspore import ops
+from mindspore import ops, version
 from mindspore.common import dtype as mstype
 from mindspore.common.tensor import Tensor
 from mindspore.nn.optim.optimizer import Optimizer, opt_init_args_register
@@ -163,6 +163,8 @@ def construct(self, gradients):
         gradients = self.scale_grad(gradients)
         gradients = self._grad_sparse_indices_deduplicate(gradients)
         lr = self.get_lr()
+        if version.__version__ >= "2.2" and self._is_dynamic_lr_or_weight_decay():
+            self.assignadd(self.global_step, self.global_step_increase_tensor)
 
         # TODO: currently not support dist
         success = self.map_(

mindocr/optim/lion.py

@@ -1,7 +1,7 @@
 import numpy as np
 
 import mindspore as ms
-from mindspore import ops
+from mindspore import ops, version
 from mindspore.common.initializer import initializer
 from mindspore.common.parameter import Parameter
 from mindspore.common.tensor import Tensor
@@ -158,6 +158,8 @@ def __init__(
     @jit
     def construct(self, gradients):
         lr = self.get_lr()
+        if version.__version__ >= "2.2" and self._is_dynamic_lr_or_weight_decay():
+            self.assignadd(self.global_step, self.global_step_increase_tensor)
         gradients = scale_grad(gradients, self.reciprocal_scale)
         if self.clip:
             gradients = ops.clip_by_global_norm(gradients, 5.0, None)

mindocr/optim/nadam.py

@@ -2,7 +2,7 @@
 import numpy as np
 
 import mindspore as ms
-from mindspore import ops
+from mindspore import ops, version
 from mindspore.common.initializer import initializer
 from mindspore.common.parameter import Parameter
 from mindspore.common.tensor import Tensor
@@ -67,6 +67,8 @@ def __init__(
     @jit
     def construct(self, gradients):
         lr = self.get_lr()
+        if version.__version__ >= "2.2" and self._is_dynamic_lr_or_weight_decay():
+            self.assignadd(self.global_step, self.global_step_increase_tensor)
         params = self.parameters
         step = self.global_step + _scaler_one
         gradients = self.decay_weight(gradients)

mindocr/utils/misc.py

@@ -35,10 +35,18 @@ def fetch_optimizer_lr(opt):
         if opt.is_group_lr:
             lr = ()
             for learning_rate in opt.learning_rate:
-                cur_dynamic_lr = learning_rate(opt.global_step - 1).reshape(())
+                cur_dynamic_lr = (
+                    learning_rate(opt.global_step - 1).reshape(())
+                    if ms.version.__version__ < "2.2"
+                    else learning_rate(opt.global_step).reshape(())
+                )
                 lr += (cur_dynamic_lr,)
         else:
-            lr = opt.learning_rate(opt.global_step - 1).reshape(())
+            lr = (
+                opt.learning_rate(opt.global_step - 1).reshape(())
+                if ms.version.__version__ < "2.2"
+                else opt.learning_rate(opt.global_step).reshape(())
+            )
     # print(f"After, global step: {opt.global_step}")
     return lr
 

tools/dataset_converters/ccpd.py

@@ -149,7 +149,7 @@ def _format_det_label(self, image_dir: Path, label_path: Path, output_path: str)
         if label_path:
             with open(label_path, "r") as f:
                 for line in f.readlines():
-                    img_path = image_dir / line.strip()
+                    img_path = image_dir / line.strip().split("/")[1]
                     assert img_path.exists(), f"Image {img_path} not found."
                     img_paths.append(img_path)
         else: