c2dconv: COCO to DOTA Annotation Format Converter

Descrtiption

Converts manual annotations created in CVAT that are exported in COCO format to Yolov5-OBB annotation format with bbox rotations.

Instructions

NOTE: This section is still WIP!

Requirements:

pip install -r requirements.txt

Usage:

c2dconv.py -i <indir>                       Convert <indir> to single dataset in ./output
c2dconv.py -i <indir> -o <outdir>           Convert <indir> to single dataset in <outdir>
c2dconv.py -i <indir> -s                    Convert <indir> to split dataset in ./output
c2dconv.py -i <indir> -o <outdir> -s        Convert <indir> to split dataset in <outdir>
c2dconv.py -z -i <indir>                    Unzip <indir> then convert to single dataset in ./output
c2dconv.py -z -i <indir> -o <outdir>        Unzip <indir> then convert to single dataset in <outdir>
c2dconv.py -z -i <indir> -s                 Unzip <indir> then convert to split dataset in ./output
c2dconv.py -z -i <indir> -o <outdir> -s     Unzip <indir> then convert to split dataset in <outdir>

Example Usage:

c2dconv.py -z -i ./Verified_with_Attributes.zip -s

Options:

-h, --help              Show this help message and exits
-z, --zip               Unzips <indir> and all inner zips before converting
-i, --indir  <indir>    Place json files and images in <indir>, cannot contain "debug" in name
-o, --outdir <outdir>   Outputs dataset to <outdir>, default is ./output, cannot contain "debug" in name
-s, --split             Splits dataset
-d, --draw              Draws rotated bounding boxes on images for debugging