Yolov5 training

projects/python/perception/object_detection_2d/yolov5/README.md

@@ -4,4 +4,7 @@ This folder contains minimal code usage examples that showcase the basic inferen
 provided by OpenDR. Specifically the following examples are provided:
 1. inference_demo.py: Perform inference on a single image. Setting `--device cpu` performs inference on CPU.
 2. webcam_demo.py: A simple tool that performs live object detection using a webcam.
-3. inference_tutorial.ipynb: Perform inference using pretrained or custom models.
+3. inference_tutorial.ipynb: Perform inference using pretrained or custom models.
+4. convert_detection_dataset.py: An example of how to convert a `DetectionDataset` into the required format
+to train a custom model. Training instructions can be found on the original 
+   [YOLOv5 repository](https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data#3-train)

projects/python/perception/object_detection_2d/yolov5/convert_detection_dataset.py

@@ -0,0 +1,112 @@
+# Copyright 2020-2023 OpenDR European Project
+#
+# 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.
+
+"""
+This example shows one way to convert a DetectionDataset, namely the AGIHumans dataset, to YOLOv5 format,
+to be used for training a custom model using the original YOLOv5 implementation: https://github.com/ultralytics/yolov5
+The opendr_datasets package can be installed using: `pip install git+https://github.com/opendr-eu/datasets.git`
+"""
+
+import argparse
+import os
+import yaml
+import cv2
+from opendr_datasets import AGIHumans
+
+
+def main(args):
+    path = args.original_data_path
+    train_set = AGIHumans(path, train=True)
+    val_set = AGIHumans(path, train=False)
+
+    new_path = args.new_data_path
+    os.makedirs(new_path, exist_ok=True)
+
+    # step 1: write dataset .yml file
+    # the new data structure is as follows:
+    # new_path
+    # ├── train
+    # │  ├── images
+    # │  │  ├── im00001.jpg
+    # │  │  └── ...
+    # │  └── labels
+    # │     ├── im00001.txt
+    # │     └── ...
+    # ├── test
+    # │  ├── images
+    # │  │  ├── im00001.jpg
+    # │  │  └── ...
+    # │  └── labels
+    # │     ├── im00001.txt
+    # │     └── ...
+    # └── AGIHumans.yml
+
+    d = {
+        'path': new_path,
+        'train': 'train',
+        'val': 'test',
+        'names': {c: c_name for c, c_name in enumerate(train_set.class_names)}
+    }
+
+    with open('AGIHumans.yaml', 'w') as yaml_file:
+        yaml.dump(d, yaml_file, default_flow_style=False)
+
+    # step 2: convert annotations to .txt files
+    # train set
+    os.makedirs(os.path.join(new_path, 'train', 'images'), exist_ok=True)
+    os.makedirs(os.path.join(new_path, 'train', 'labels'), exist_ok=True)
+    for idx, (img, boxes) in enumerate(train_set):
+        # save img to 'train/images/im{:05d}.jpg'
+        im_cv = img.opencv()
+        cv2.imwrite(os.path.join(new_path, 'train', 'images', f'im{idx:05d}.jpg'), im_cv)
+        im_height, im_width, im_c = im_cv.shape
+        # save normalized label to 'train/labels/im{:05d}.txt
+        lines = ''
+        for box in boxes:
+            x_center = (box.left + box.width * 0.5) / im_width
+            y_center = (box.top + box.height * 0.5) / im_height
+            width = box.width / im_width
+            height = box.height / im_height
+            lines += f'{box.name} {x_center} {y_center} {width} {height}\n'
+        if len(lines) > 0:
+            with open(os.path.join(new_path, 'train', 'labels', f'im{idx:05d}.txt'), 'w') as f:
+                f.write(lines)
+
+    # validation/test set
+    os.makedirs(os.path.join(new_path, 'test', 'images'), exist_ok=True)
+    os.makedirs(os.path.join(new_path, 'test', 'labels'), exist_ok=True)
+    for idx, (img, boxes) in enumerate(val_set):
+        # save img to 'train/images/im{:05d}.jpg'
+        im_cv = img.opencv()
+        cv2.imwrite(os.path.join(new_path, 'test', 'images', f'im{idx:05d}.jpg'), im_cv)
+        im_height, im_width, im_c = im_cv.shape
+        # save normalized label to 'train/labels/im{:05d}.txt
+        lines = ''
+        for box in boxes:
+            x_center = (box.left + box.width * 0.5) / im_width
+            y_center = (box.top + box.height * 0.5) / im_height
+            width = box.width / im_width
+            height = box.height / im_height
+            lines += f'{box.name} {x_center} {y_center} {width} {height}\n'
+        if len(lines) > 0:
+            with open(os.path.join(new_path, 'test', 'labels', f'im{idx:05d}.txt'), 'w') as f:
+                f.write(lines)
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--original-data-path", help="Dataset root", type=str)
+    parser.add_argument("--new-data-path", help="Path to converted dataset location", type=str)
+
+    args = parser.parse_args()

src/opendr/perception/object_detection_2d/yolov5/yolov5_learner.py

@@ -40,7 +40,7 @@ def __init__(self, model_name, path=None, device='cuda', temp_path='.', force_re
                                         force_reload=force_reload)
         else:
             self.model = torch.hub.load('ultralytics/yolov5:master', 'custom', path=path,
-                                        force_reload=force_reload, skip_validation=True)
+                                        force_reload=force_reload)
         torch.hub.set_dir(default_dir)
 
         self.model.to(device)