YOLO-Coco-Dataset-Custom-Classes-Extractor

• Download specific classes from the Coco Dataset for custrom object detection needs.
• Download multiple classes at the same time (Multi-threaded).
• Pickup where you left off if your connection is interrupted.

Packages Required

1. pycocotools
pip install pycocotools

Usage

1. Clone this repository:

git clone https://github.com/KaranJagtiani/YOLO-Coco-Dataset-Custom-Classes-Extractor.git

2. Download the 2017 Train/Val annotations [241MB] zip file and put the instances_train2017.json file in the cloned repository's main directory.

3. See the various classes available:

python coco-extractor.py --help

4. Download a specific class:

python coco-extractor.py "person"

5. Download multiple classes:

python coco-extractor.py "person" "sports ball" "zebra"

Test your downloaded images and the bounding boxes with labelImg.

Happy Detecting!