metadata
license: mit
Dataset for clothes segmentation
Based on iMaterialist Dataset with several adjustments:
- Filtered images with more than 1 person with no labeled clothes
- Compressed number of classes to 8:
'background', 'upperbody', 'upperbody_up', 'lowerbody', 'wholebody', 'wholebody_up', 'shoes', 'accesories' - Simple structure with 2 folders: images 512x512 in
.jpgand corresponding segmaps 512x512 in.npy - You can find example class and data vizualisation in
dataset.ipynb
If you find any bugs, please contact me on: [email protected]