In addition, performance-wise it gets closer to state-of-the-art supervised methods such as Mask R-CNN, getting closer to learning image representation as humans do. Writing for InfoQ has opened ...
