revise the script and remove -split indicator

docs/en/user_guides/useful_tools.md

@@ -135,12 +135,16 @@ Instructions:
 └── $ROOT_PATH
          ├── class.txt
          ├── labels
+             ├── a.txt
+             ├── b.txt
          ├── images
+             ├── a.jpg
+             ├── b.png
          └── ...
 ```
 
-2. `split` indicates whether you need to format the dataset based on the existing `train`, `val`, and `text` categories. It is not used by default. If you need to use it, please ensure the corresponding `train.txt`, ` val.txt`, and `test.txt` must exist under `image-dir`. Otherwise, the script will fail to run.
-3. By default, the script will create a folder called `coco_format` in the `image-dir` directory where stores the converted JSON file. If `split` is not specified, the output file is `result.json`. If `split` is specified, three corresponding files called `train.json`, `val.json`, and `test.json` will be generated.
+2. The script will automatically check if `train.txt`, `val.txt`, and `test.txt` are already existed under `image-dir`. If these files are located, the script will organize the dataset accordingly. Otherwise, the script will convert the dataset in one file.
+3. By default, the script will create a folder called `coco_format` in the `image-dir` directory where stores the converted JSON file. If `train.txt`, `val.txt`, and `test.txt` are not found, the output file is `result.json`. Otherwise, three corresponding files called `train.json`, `val.json`, and `test.json` will be generated.
 
 ## Download Dataset
 

docs/zh_cn/user_guides/useful_tools.md

@@ -128,20 +128,24 @@ python tools/dataset_converters/yolo2coco.py --image-dir /path/to/the/root/dir/o
 
 使用说明：
 
-1. `image-dir` 是待转换的yolo格式数据集的根目录，内应包含 `images` 、 `labels` 和 `class.txt` 文件， `class.txt` 是当前dataset对应的类的声明，一行一个。如下所示：
-
+1. `image-dir` 是待转换的yolo格式数据集的根目录，内应包含 `images` 、 `labels` 和 `class.txt` 文件， `class.txt` 是当前dataset对应的类的声明，一行一个类别。
+   `image_dir` 结构如下例所示：
 
 ```bash
 .
 └── $ROOT_PATH
          ├── class.txt
          ├── labels
+             ├── a.txt
+             ├── b.txt
          ├── images
+             ├── a.jpg
+             ├── b.png
          └── ...
 ```
 
-2. `split` 代表是否需要在转换中按已有的 `train` 、 `val` 和 `text` 类别进行整理，默认不使用，如果需要使用，请确保对应的 `train.txt` 、 `val.txt` 和 `test.txt` 也要存在于 `image-dir` 下，否则将报错。
-3. 脚本会默认在 `image-dir` 目录下创建 `coco_format` 文件夹并将转换结果存在这里。如果未指明 `split` ，输出文件即为 `result.json`，如果指明 `split` ，则会生成对应的 `train.json` 、 `val.json` 和 `test.json` 三个文件。
+2. 脚本会检测 `image-dir` 下是否已有的 `train.txt` 、 `val.txt` 和 `test.txt` 。若检测到三个文件，则会按照类别进行整理， 否则默认不需要分类。故请确保对应的 `train.txt` 、 `val.txt` 和 `test.txt` 要在 `image-dir` 内。
+3. 脚本会默认在 `image-dir` 目录下创建 `coco_format` 文件夹并将转换结果存在这里。如果在 `image-dir` 下没找到分类文件，输出文件即为 `result.json`，反之，则会生成对应的 `train.json` 、 `val.json` 和 `test.json` 三个文件。
 
 ## 数据集下载
 

tools/dataset_converters/yolo2coco.py

@@ -2,20 +2,23 @@
 
 Usage:
     $ python yolo2coco.py --image-dir /path/to/dataset   # the root dir
-                          [--split]                      # if splits
 
 Note:
     1. Before running this script, please make sure the root directory
     of your dataset is formatted in the following struction:
     .
     └── $ROOT_PATH
-             ├── class.txt
-             ├── labels
-             ├── images
-             └── ...
-    2. `split` is not used by default. If you need to use it, please ensure
-    the corresponding`train.txt`, ` val.txt`, and `test.txt` must exist under
-    `image-dir`. Otherwise, the script will fail to run.
+            ├── class.txt
+            ├── labels
+                ├── a.txt
+                ├── b.txt
+            ├── images
+                ├── a.jpg
+                ├── b.png
+            └── ...
+    2. The script will automatically check whether the corresponding
+    `train.txt`, ` val.txt`, and `test.txt` exist under `image-dir` or not.
+    If these files are detected, the script will organize the dataset.
     3. Once the script finishes, the result files will be saved in the
     directory named 'coco_format' in the root directory of your dataset.
 """
@@ -96,14 +99,12 @@ def organize_by_existing_files(image_dir: str):
     return image_list[0], image_list[1], image_list[2]
 
 
-def convert_yolo_to_coco(image_dir: str, split: bool = False):
+def convert_yolo_to_coco(image_dir: str):
     """Convert annotations from yolo style to coco style.
 
     Args:
         image_dir (str): the root directory of your datasets which contains
             labels, images, classes.txt, etc
-        split (bool): whether to organize the datasets based on existing
-            train.txt, val.txt, and test.txt
     """
     print(f'Start to load existing images and annotations from {image_dir}')
     check_existance(image_dir)
@@ -114,14 +115,27 @@ def convert_yolo_to_coco(image_dir: str, split: bool = False):
     check_existance(yolo_label_dir)
     check_existance(yolo_image_dir)
     check_existance(yolo_class_txt)
+    print(f'All necessary files are located at {image_dir}')
+
+    train_txt_path = osp.join(image_dir, 'train.txt')
+    val_txt_path = osp.join(image_dir, 'val.txt')
+    test_txt_path = osp.join(image_dir, 'test.txt')
+    print(f'Checking if train.txt, val.txt, and test.txt are in {image_dir}')
+    if osp.exists(train_txt_path) and osp.exists(val_txt_path) and osp.exists(
+            test_txt_path):
+        print(
+            'these files are located, need to organize the data accordingly.')
+        to_categorize = True
+    else:
+        print('these files are not located, no need to organize separately.')
+        to_categorize = False
 
     with open(yolo_class_txt) as f:
         classes = f.read().strip().split()
 
     indices = os.listdir(yolo_image_dir)
 
-    if split:
-        print('Start to work based to existing train, test, and val')
+    if to_categorize:
         train_dataset = {'images': [], 'annotations': [], 'categories': []}
         val_dataset = {'images': [], 'annotations': [], 'categories': []}
         test_dataset = {'images': [], 'annotations': [], 'categories': []}
@@ -133,7 +147,6 @@ def convert_yolo_to_coco(image_dir: str, split: bool = False):
             test_dataset['categories'].append({'id': i, 'name': cls})
         train_img, val_img, test_img = organize_by_existing_files(image_dir)
     else:
-        print('Start to work on all data')
         dataset = {'images': [], 'annotations': [], 'categories': []}
         for i, cls in enumerate(classes, 0):
             dataset['categories'].append({'id': i, 'name': cls})
@@ -148,7 +161,7 @@ def convert_yolo_to_coco(image_dir: str, split: bool = False):
         img_name = osp.splitext(image)[0]
         img_info_dict, H, W = get_image_info(yolo_image_dir, idx, image)
 
-        if split:
+        if to_categorize:
             if image in train_img:
                 dataset = train_dataset
             elif image in val_img:
@@ -177,7 +190,7 @@ def convert_yolo_to_coco(image_dir: str, split: bool = False):
     if not osp.exists(res_folder):
         os.makedirs(res_folder)
 
-    if split:
+    if to_categorize:
         for category in ['train', 'val', 'test']:
             out_file = osp.join(image_dir, f'coco_format/{category}.json')
             print(f'Saving converted annotations to {out_file}')
@@ -201,9 +214,5 @@ def convert_yolo_to_coco(image_dir: str, split: bool = False):
         type=str,
         required=True,
         help='dataset directory with ./images and ./labels, classes.txt, etc.')
-    parser.add_argument(
-        '--split',
-        action='store_true',
-        help='convert based on existing train.txt, val.txt, and test.txt')
     arg = parser.parse_args()
-    convert_yolo_to_coco(arg.image_dir, arg.split)
+    convert_yolo_to_coco(arg.image_dir)