CVAT Segmentation to YOLOv8 Converter

A Python script that converts CVAT polygons / segmentation data / annotations in COCO JSON format to YOLOv8 compatible format for object detection and segmentation tasks.

Features

• Converts COCO JSON format to YOLOv8 annotation format
• Handles polygon segmentation data
• Normalizes coordinates according to YOLOv8 requirements
• Preserves category IDs and mappings
• Processes multiple images and annotations

Prerequisites

• Python 3.6 or higher
• Access to CVAT-exported annotations in COCO JSON format

Installation

Clone this repository:

git clone https://github.com/SakibAhmedShuva/cvat-yolov8-converter.git
cd cvat-yolov8-converter

Usage

1. Update the coco_file path in the script to point to your COCO JSON file:

coco_file = '/path/to/your/instances_default.json'

2. Run the script:

python cvat_to_yolov8.py

The script will:

• Create a directory named yolov8_annotations (if it doesn't exist)
• Process each annotation in the COCO JSON file
• Generate corresponding YOLOv8 format text files for each image

Output Format

For each image in the COCO dataset, the script creates a text file with the same name as the image (but with .txt extension) containing the annotations in YOLOv8 format:

<category_id> <x1> <y1> <x2> <y2> ... <xn> <yn>

Where:

• category_id is the class ID
• x1, y1, x2, y2, ..., xn, yn are normalized polygon coordinates

Example

Input (COCO format):

{
  "images": [{"id": 1, "width": 800, "height": 600, "file_name": "image1.jpg"}],
  "annotations": [
    {
      "image_id": 1,
      "category_id": 0,
      "segmentation": [[100, 100, 200, 100, 200, 200, 100, 200]]
    }
  ]
}

Output (YOLOv8 format in image1.txt):

0 0.125 0.166667 0.25 0.166667 0.25 0.333333 0.125 0.333333

Limitations

• Currently only supports polygon segmentation
• Assumes one segmentation per annotation
• Does not handle complex, nested polygons

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

https://github.com/SakibAhmedShuva

Project Link: https://github.com/SakibAhmedShuva/CVAT-Segmentation-to-YOLOv8-Converter