Object detection support

Hello, great work!
Is it possible to extend this to object detection tasks? Is that something which was pursued during this research but hasn't been included in the paper/repo due to some reason? Thanks.