[Feature] Support training on a single channel image dataset.

projects/single_channel/README.md

@@ -0,0 +1,35 @@
+# MMYOLO Model #409
+
+## Introduction
+
+This project was developed for training on a single channel image dataset.
+
+Now, the script only support `YOLOv5` .
+
+## Usage
+
+### Data set preparation
+
+If you are using a custom grayscale image dataset, you can skip this step.
+
+```shell
+python tools/misc/download_dataset.py --dataset-name balloon --save-dir projects/single_channel/data --unzip
+python projects/single_channel/balloon2coco_single_channel.py
+python projects/single_channel/single_channel.py
+```
+
+### Training
+
+In the configuration file, `_base_` and `data_root` need to be modified to the corresponding dataset paths.
+
+```shell
+python tools/train.py projects/single_channel/configs/yolov5_s-v61_syncbn_fast_1xb4-300e_balloon_single_channel.py
+```
+
+### Model Testing
+
+```shell
+python tools/test.py projects/single_channel/configs/yolov5/yolov5_s-v61_syncbn_fast_1xb4-300e_balloon.py \
+                     work_dirs/yolov5_s-v61_syncbn_fast_1xb4-300e_balloon/epoch_300.pth \
+                     --show-dir show_results
+```

projects/single_channel/balloon2coco_single_channel.py

@@ -0,0 +1,64 @@
+import os.path as osp
+
+import mmcv
+import mmengine
+
+
+def convert_balloon_to_coco(ann_file, out_file, image_prefix):
+
+    data_infos = mmengine.load(ann_file)
+
+    annotations = []
+    images = []
+    obj_count = 0
+    for idx, v in enumerate(mmengine.track_iter_progress(data_infos.values())):
+        filename = v['filename']
+        img_path = osp.join(image_prefix, filename)
+        height, width = mmcv.imread(img_path).shape[:2]
+
+        images.append(
+            dict(id=idx, file_name=filename, height=height, width=width))
+
+        for _, obj in v['regions'].items():
+            assert not obj['region_attributes']
+            obj = obj['shape_attributes']
+            px = obj['all_points_x']
+            py = obj['all_points_y']
+            poly = [(x + 0.5, y + 0.5) for x, y in zip(px, py)]
+            poly = [p for x in poly for p in x]
+
+            x_min, y_min, x_max, y_max = (min(px), min(py), max(px), max(py))
+
+            data_anno = dict(
+                image_id=idx,
+                id=obj_count,
+                category_id=0,
+                bbox=[x_min, y_min, x_max - x_min, y_max - y_min],
+                area=(x_max - x_min) * (y_max - y_min),
+                segmentation=[poly],
+                iscrowd=0)
+            annotations.append(data_anno)
+            obj_count += 1
+
+    coco_format_json = dict(
+        images=images,
+        annotations=annotations,
+        categories=[{
+            'id': 0,
+            'name': 'balloon'
+        }])
+    mmengine.dump(coco_format_json, out_file)
+
+
+if __name__ == '__main__':
+
+    convert_balloon_to_coco(
+        'projects/single_channel/data/balloon/train/'
+        'via_region_data.json',
+        'projects/single_channel/data/balloon/train.json',
+        'projects/single_channel/data/balloon/train/')
+    convert_balloon_to_coco(
+        'projects/single_channel/data/balloon/val/'
+        'via_region_data.json',
+        'projects/single_channel/data/balloon/val.json',
+        'projects/single_channel/data/balloon/val/')

projects/single_channel/configs/yolov5_s-v61_syncbn_fast_1xb4-300e_balloon_single_channel.py

@@ -0,0 +1,39 @@
+_base_ = '../../../configs/yolov5/yolov5_s-v61_syncbn_fast_8xb16-300e_coco.py'
+
+data_root = '../../../projects/single_channel/data/balloon/'
+
+train_batch_size_per_gpu = 4
+train_num_workers = 2
+
+metainfo = {
+    'classes': ('balloon', ),
+    'palette': [
+        (220, 20, 60),
+    ]
+}
+
+train_dataloader = dict(
+    batch_size=train_batch_size_per_gpu,
+    num_workers=train_num_workers,
+    dataset=dict(
+        data_root=data_root,
+        metainfo=metainfo,
+        data_prefix=dict(img='train/'),
+        ann_file='train.json'))
+
+val_dataloader = dict(
+    dataset=dict(
+        data_root=data_root,
+        metainfo=metainfo,
+        data_prefix=dict(img='val/'),
+        ann_file='val.json'))
+
+test_dataloader = val_dataloader
+
+val_evaluator = dict(ann_file=data_root + 'val.json')
+
+test_evaluator = val_evaluator
+
+model = dict(bbox_head=dict(head_module=dict(num_classes=1)))
+
+default_hooks = dict(logger=dict(interval=1))

projects/single_channel/single_channel.py

@@ -0,0 +1,29 @@
+import imghdr
+import os
+from typing import List
+
+from PIL import Image
+
+path = 'projects/single_channel/data/balloon/train'
+save_path = 'projects/single_channel/data/balloon/train'
+file_list: List[str] = os.listdir(path)
+
+# Grayscale conversion of each image
+for file in file_list:
+    if imghdr.what(path + '/' + file) != 'jpeg':
+        continue
+    o_img = Image.open(path + '/' + file)
+    L_img = o_img.convert('L')
+    L_img.save(save_path + '/' + file)
+
+path = 'projects/single_channel/data/balloon/val'
+save_path = 'projects/single_channel/data/balloon/val'
+file_list = os.listdir(path)
+
+# Grayscale conversion of each image
+for file in file_list:
+    if imghdr.what(path + '/' + file) != 'jpeg':
+        continue
+    o_img = Image.open(path + '/' + file)
+    L_img = o_img.convert('L')
+    L_img.save(save_path + '/' + file)