DeepFashion2 is a comprehensive fashion dataset. It contains 491K diverse images of 13 popular clothing categories from both commercial shopping stores and consumers. It totally has 801K clothing clothing items, where each item in an image is labeled with scale, occlusion, zoom-in, viewpoint, category, style, bounding box, dense landmarks and per-pixel mask.There are also 873K Commercial-Consumer clothes pairs. The full dataset can be downloaded with the instruction at - https://github.com/switchablenorms/DeepFashion2.
Mask R-CNN was trained with this dataset to segement and seperate different fashion items of the fashion set. This is of very high significance to anyone researching in the field of fashion recommendation.
Example segementation -
