How do Networks predict amodal masks

[staple-pi]

trafficstars

Hi, I have read your paper. And i have noticed that in the Experiments part, you evaluate the performance of BCNets on amodal dataset like cocoa. With figure4 i know that BCNet can predict modal masks of occluder and occludee, but i confuse how the BCNet predict amodal masks like figure5. Does it need any post-processing？

[lkeab]

hi, sorry for the late reply, for cocoa, it need some processing on annotation file, similar to here