Python scripts for converting instance maps/class tags to RGB semantic segmentation maps and vice-versa.
rgb.py: converts instance image maps to RGB segmentation maps.
inst.py: converts RGB segmentation maps to instance image maps.
'color' is a list of color labels for each class numerically arranged w.r.t each class number tag. Modify this in the scripts depending your requirements.
video.py: converts image frames to a video sequence at a specified fps rate.
Install python packages from requirements.txt
pip install -r requirements.txt
Instance maps to RGB maps
python rgb.py --path <datapath> --imagewidth <size> --imageheight <size> --savepath <datapath>
RGB Maps to Instance maps
python inst.py --path <datapath> --imagewidth <size> --imageheight <size> --savepath <datapath>
By default, the converted images are saved in './save/' folder and image size is 2048x1024.
Disclaimer: You will not receive the same color masks if you try to convert back and forth. I think it's because the colors are not properly interpreted during this. Feel free to fix this and leave a PR.
Image Frames to Video Sequence
python video.py --path <datapath> --fps <video fps>
Stores as video.avi in the same directory of script.