COCO to Yolo Downloader (Images, Annoatations, Negatives)

Downloads COCO dataset by multiple image categories in parallel threads, converts COCO annotations to YOLO format and stored in respective .txt files
Download Negative images which excludes the categories in categories_to_download.txt
Can include custom class numbers to be added to annoation, just add desired numbers to categories_to_download.txt
Can limit downloads based on per class

Requirements:

pip3 install pycococtools
COCO Dataset "instances_train2017.json" file from here

Run:

Update categories_to_download.txt file with category names and calss numbers that you want to add.
To download coco dataset with limit of 5000 images per class:

 python3 main.py -o download -l 5000

To download negatives from coco dataset with limit of 1000:

python3 main.py -o negatives -l 1000

By default annotations along with images are stored in annotations folder, you can change it here
Image download will be skipped if already exists. However, annotations will always be appended to the existing file so it's advisable to delete *.txt file before a re-run
Do not delete or modify coco-names.txt