Convert LabelMe to MsCOCO, PascalVOC, Yolo format. As well as splitting the dataset into training, testing, and validation samples; augmenting the transformed dataset with data augmentation; outputting and recording image statistics; and converting the entire dataset to a single resolution.

Installing the necessary packages
>> pip install -r requirements.txt

The list of classes is written in the .txt file in the order they appear in the statistics (by default in labels.txt). The name of the filename is specified in config.py.

with parameterslabelme_converter.pyFiles used for conversion:

• --input– Directory containing datasets in LabelMe format
• --output– (optional) directory to store the transformed dataset in (default./current_dataset🙂
• --format– Conversion format ( yolo, voc, coco)
• --create-marked -(
• --poly– (optional) specifies whether the visualization should use polygons (instead of rectangles)

Example:

Converting dataset from ./meters01_labelme to Yolo format with creation of markup visualization
>> python labelme_converter.py --input meters01_labelme --output current_dataset --format yolo --create-marked

Input path:  labelme
Output path:  current_dataset
Reading labelme files:
100%|██████████████████████████████████████████| 24/24 [00:00<00:00, 98.69it/s]
Create marked images:
100%|██████████████████████████████████████████| 24/24 [00:00<00:00, 50.44it/s]
Converting labelme to Yolo:
100%|██████████████████████████████████████████| 24/24 [00:00<00:00, 8058.87it/s]