Add the backgroud images to the train dataset.

Author: SWHL

README.md

@@ -1,9 +1,13 @@
-#### YOLOV5格式数据→COCO
-- 值得一提的是，由标注软件[labelImg](https://github.com/tzutalin/labelImg)标注所得yolo格式数据，也可由该脚本做转换。前提是满足以下数据目录结构。
+## YOLO2COCO
+
+#### YOLOV5格式数据 → COCO
+- 可以将一些相似的背景图像加入到训练中，直接将背景图像放入`backgroud_images`目录即可，转换程序会自动扫描该目录，添加到训练集中。
 - YOLOV5训练格式目录结构（详情参见`dataset/YOLOV5`）：
     ```text
     YOLOV5
     ├── classes.txt
+    ├── background_images  # 背景图像，一般是和要检测的对象容易混淆的图像
+    │   └── bg1.jpeg
     ├── images
     │   ├── images(13).jpg
     │   └── images(3).jpg
@@ -15,8 +19,10 @@
     ```
 
 - 转换
+  - `--dir_path`：整理好的数据集所在目录
+  - `--mode_list`：指定生成的json类型，前提是要有对应的txt文件，可单独指定。
   ```shell
-  python yolov5_2_coco.py --dir_path dataset/YOLOV5
+  python yolov5_2_coco.py --dir_path dataset/YOLOV5 --mode_list train,val
   ```
 - 转换后目录结构（详情参见`dataset/YOLOV5_COCO_format`）：
     ```text
@@ -25,12 +31,13 @@
     │   ├── instances_train2017.json
     │   └── instances_val2017.json
     ├── train2017
-    │   └── 000000000001.jpg
+    │   ├── 000000000001.jpg
+    │   └── 000000000002.jpg  # 这个是背景图像
     └── val2017
         └── 000000000001.jpg
     ```
 
-#### YOLOV5 yaml数据描述文件转→COCO
+#### YOLOV5 yaml数据描述文件转 → COCO
 
 - YOLOV5 yaml 数据文件需要包含：
     ```text
@@ -57,7 +64,7 @@
   python yolov5_yaml_2_coco.py --yaml_path dataset/YOLOV5_yaml/sample.yaml
   ```
 
-#### darknet格式数据→COCO
+#### darknet格式数据 → COCO
 - darknet训练数据目录结构（详情参见`dataset/darknet`）：
   ```text
   darknet

coco_visual.py

@@ -15,13 +15,13 @@ def visualization_bbox(num_image, json_path, img_path):
         annotation_json = json.load(annos)
 
     # 统计json文件的关键字长度
-    print('the annotation_json num_key is:', len(annotation_json))
+    print('The annotation_json num_key is:', len(annotation_json))
 
     # 读出json文件的关键字
-    print('the annotation_json key is:', annotation_json.keys())
+    print('The annotation_json key is:', annotation_json.keys())
 
     # json文件中包含的图片数量
-    print('the annotation_json num_images is:', len(annotation_json['images']))
+    print('The annotation_json num_images is:', len(annotation_json['images']))
 
     # 获取所有类别数
     categories = annotation_json['categories']

dataset/YOLOV5/background_images/bg1.jpeg

@@ -1 +1 @@
-{"info": {"year": 2021, "version": "1.0", "description": "For object detection", "date_created": "2021"}, "images": [{"date_captured": "2021", "file_name": "000000000001.jpg", "id": 1, "height": 224, "width": 224}], "licenses": [{"id": 1, "name": "Apache License v2.0", "url": "https://github.com/RapidAI/YOLO2COCO/LICENSE"}], "type": "instances", "annotations": [{"segmentation": [[18.000080000000004, 2.999920000000003, 105.000112, 2.999920000000003, 105.000112, 89.000016, 18.000080000000004, 89.000016]], "area": 7482.011104003072, "iscrowd": 0, "image_id": 1, "bbox": [18.000080000000004, 2.999920000000003, 87.000032, 86.000096], "category_id": 1, "id": 1}, {"segmentation": [[126.99993599999999, 3.9999679999999955, 210.999936, 3.9999679999999955, 210.999936, 88.999904, 126.99993599999999, 88.999904]], "area": 7139.994624000001, "iscrowd": 0, "image_id": 1, "bbox": [126.99993599999999, 3.9999679999999955, 84.0, 84.999936], "category_id": 1, "id": 2}, {"segmentation": [[74.99990400000002, 76.000064, 150.99996800000002, 76.000064, 150.99996800000002, 153.000064, 74.99990400000002, 153.000064]], "area": 5852.004927999999, "iscrowd": 0, "image_id": 1, "bbox": [74.99990400000002, 76.000064, 76.000064, 77.0], "category_id": 1, "id": 3}, {"segmentation": [[12.000127999999997, 135.000096, 97.00006400000001, 135.000096, 97.00006400000001, 220.00003200000003, 12.000127999999997, 220.00003200000003]], "area": 7224.989120004097, "iscrowd": 0, "image_id": 1, "bbox": [12.000127999999997, 135.000096, 84.999936, 84.999936], "category_id": 1, "id": 4}, {"segmentation": [[132.00006399999998, 136.999968, 215.00012799999996, 136.999968, 215.00012799999996, 220.00003199999998, 132.00006399999998, 220.00003199999998]], "area": 6889.010624004095, "iscrowd": 0, "image_id": 1, "bbox": [132.00006399999998, 136.999968, 83.000064, 83.000064], "category_id": 1, "id": 5}], "categories": [{"supercategory": "stamp", "id": 1, "name": "stamp"}]}
+{"info": {"year": 2022, "version": "1.0", "description": "For object detection", "date_created": "2022"}, "images": [{"date_captured": "2022", "file_name": "000000000001.jpg", "id": 1, "height": 224, "width": 224}, {"date_captured": "2022", "file_name": "000000000002.jpg", "id": 2, "height": 424, "width": 550}], "licenses": [{"id": 1, "name": "Apache License v2.0", "url": "https://github.com/RapidAI/YOLO2COCO/LICENSE"}], "type": "instances", "annotations": [{"segmentation": [[18.000080000000004, 2.999920000000003, 105.000112, 2.999920000000003, 105.000112, 89.000016, 18.000080000000004, 89.000016]], "area": 7482.011104003072, "iscrowd": 0, "image_id": 1, "bbox": [18.000080000000004, 2.999920000000003, 87.000032, 86.000096], "category_id": 1, "id": 1}, {"segmentation": [[126.99993599999999, 3.9999679999999955, 210.999936, 3.9999679999999955, 210.999936, 88.999904, 126.99993599999999, 88.999904]], "area": 7139.994624000001, "iscrowd": 0, "image_id": 1, "bbox": [126.99993599999999, 3.9999679999999955, 84.0, 84.999936], "category_id": 1, "id": 2}, {"segmentation": [[74.99990400000002, 76.000064, 150.99996800000002, 76.000064, 150.99996800000002, 153.000064, 74.99990400000002, 153.000064]], "area": 5852.004927999999, "iscrowd": 0, "image_id": 1, "bbox": [74.99990400000002, 76.000064, 76.000064, 77.0], "category_id": 1, "id": 3}, {"segmentation": [[12.000127999999997, 135.000096, 97.00006400000001, 135.000096, 97.00006400000001, 220.00003200000003, 12.000127999999997, 220.00003200000003]], "area": 7224.989120004097, "iscrowd": 0, "image_id": 1, "bbox": [12.000127999999997, 135.000096, 84.999936, 84.999936], "category_id": 1, "id": 4}, {"segmentation": [[132.00006399999998, 136.999968, 215.00012799999996, 136.999968, 215.00012799999996, 220.00003199999998, 132.00006399999998, 220.00003199999998]], "area": 6889.010624004095, "iscrowd": 0, "image_id": 1, "bbox": [132.00006399999998, 136.999968, 83.000064, 83.000064], "category_id": 1, "id": 5}, {"segmentation": [], "area": 0, "iscrowd": 0, "image_id": 2, "bbox": [], "category_id": -1, "id": 6}], "categories": [{"supercategory": "stamp", "id": 1, "name": "stamp"}]}

dataset/YOLOV5_COCO_format/annotations/instances_train2017.json

@@ -1 +1 @@
-{"info": {"year": 2021, "version": "1.0", "description": "For object detection", "date_created": "2021"}, "images": [{"date_captured": "2021", "file_name": "000000000001.jpg", "id": 1, "height": 248, "width": 203}], "licenses": [{"id": 1, "name": "Apache License v2.0", "url": "https://github.com/RapidAI/YOLO2COCO/LICENSE"}], "type": "instances", "annotations": [{"segmentation": [[18.999988000000002, 16.999904, 183.000034, 16.999904, 183.000034, 178.999952, 18.999988000000002, 178.999952]], "area": 26568.015324002205, "iscrowd": 0, "image_id": 1, "bbox": [18.999988000000002, 16.999904, 164.000046, 162.000048], "category_id": 1, "id": 6}], "categories": [{"supercategory": "stamp", "id": 1, "name": "stamp"}]}
+{"info": {"year": 2022, "version": "1.0", "description": "For object detection", "date_created": "2022"}, "images": [{"date_captured": "2022", "file_name": "000000000001.jpg", "id": 1, "height": 248, "width": 203}], "licenses": [{"id": 1, "name": "Apache License v2.0", "url": "https://github.com/RapidAI/YOLO2COCO/LICENSE"}], "type": "instances", "annotations": [{"segmentation": [[18.999988000000002, 16.999904, 183.000034, 16.999904, 183.000034, 178.999952, 18.999988000000002, 178.999952]], "area": 26568.015324002205, "iscrowd": 0, "image_id": 1, "bbox": [18.999988000000002, 16.999904, 164.000046, 162.000048], "category_id": 1, "id": 7}], "categories": [{"supercategory": "stamp", "id": 1, "name": "stamp"}]}