search

ucas-vg / P2BNet

ECCV2022, Point-to-Box Network for Accurate Object Detection via Single Point Supervision
MIT License
63 stars 10 forks source link

How to train on one GPU #6

Closed cxq1 closed 1 year ago

cxq1 commented 2 years ago

I set the batch size to 8 on one gpu, how should I set the learning rate When I set the learning rate to 0.01, the ap value on coco2017 is only 20.0

Darren-pfchen commented 1 year ago

batch size 8 for one GPU is high beyond the memory, and one GPU is time-consuming. We didn't try to do this on one GPU. If you implement it on one stage, you can report it for me. Thank you!