For a retail blockchain startup, we were tasked with building a predictive model and interface to segment different clothing items from images of humans. This segmentation was used to provide more data to the underlying reccomendation system they were building. We utilized a system to hand-draw and label training data, then leveraged a MaskRCNN model to learn the clothing types and segment them from the rest of the human images. The segmented clothing images were then outputted to the larger database of individual clothing items to be used for completing the engine.
