1 paper
Alex Foo, Wynne Hsu, Mong Li Lee
Discovering object-centric representations from images can significantly enhance the robustness, sample efficiency and generalizability of vision models. Works on images with multi…