ML_COCO_extract_specific_classes
Extract images and annotations from COCO for specific classes
Table of contents
Explanation
This code basically extracts images and its annotations from COCO dataset for specific classes/categories. The output is formatted to work with Darknet branch by AlexeyAB. This script helps with steps 4 and 5 in How to train (to detect your custom objects) in AlexeyAB repo
However, with some tweaks you can get the desired for format for your own dataset as I considered making the code as scalable and modular as possible.
Prerequisites
- COCO dataset downloaded. use this guide
- Darknet branch by AlexeyAB.
Running the code
change these lines to folder location
COCO_FOLDER = "/path/to/coco/"
DARKNET_FOLDER = "/path/to/darknet/"change the list to the desired class names
cat_names = ['car', 'bus', 'truck', 'traffic light' , 'stop sign', 'parking meter']