hasty.ai COCO2YOLO

So you wanna convert your hasty.ai's COCO format to ultralytic's YOLO dataset format, you've come to the right place! This repository is based of the ultralytics JSON2YOLO repository but is customized for exporting from hasty.ai.

Usage

Before using this script, it's important to understand the folder structures we're converting from and into.

File Structures

COCO Format (Input)

❗ Note that the images folder requires the train and val folders inside with the relevant images you exported.

├── <annotations-file>.json
├── images
    ├── train
        ├── <image1>.jpg
        ├── <image2>.jpg
        └── <image3>.jpg
    └── val
        ├── <image4>.jpg
        ├── <image5>.jpg
        └── <image6>.jpg

YOLO Format (Output)

├── <output-directory-name>
    ├── <output-directory-name>.yaml
    ├── images
        ├── train
            ├── <image1>.jpg
            ├── <image2>.jpg
            └── <image3>.jpg
        └── val
            ├── <image4>.jpg
            ├── <image5>.jpg
            └── <image6>.jpg
    ├── labels
        ├── train
            ├── <image1>.txt
            ├── <image2>.txt
            └── <image3>.txt
        └── val
            ├── <image4>.txt
            ├── <image5>.txt
            └── <image6>.txt

Example Usage

Run general_json2yolo.py with the following arguments:

-j : (required) json directory
-i : (required) images directory (with <images>/train and <images>/val)
-o : (required) output directory
-k : (optional) ids of classes to keep based on the list below
-z : (optional) enables zipping of the folder for one less step so that you can upload it to ultralytics immediately

For example, running this below:

python3 general_json2yolo.py -j "/home/samuel/x.json" -i "/home/samuel/images" -o /home/samuel/output -k 0 1 2 -z

✔️ chooses ids 0, 1, 2 based on the yaml detailing the label ids.