We recently developed a solution based on deep learning, i.e. Tiramisu, to segment the forground and remove the background in images. Our goal was to enable image retreival from a target catalogue image of a fashion retailer given a query mobile snapshot.
In this repository, you'll find:
-
the code to reproduce the image segmetnation described in the accompanying code story
-
how to train your own specific image segmentation with your own data
A trained model can be found here.