how to modify code to support image_size>1 when training?

msracver / Relation-Networks-for-Object-Detection

Relation Networks for Object Detection

MIT License

1.09k stars 190 forks source link

Closed jiangzhengkai closed 6 years ago

chengdazhi commented 6 years ago

Hi, right now the ProposalTarget custom operator only supports single image each GPU, so that's where you might want to start.

jiangzhengkai commented 6 years ago

I'm trying to modify op, thanks!

chengdazhi commented 6 years ago

No bother, good luck!