mx-maskrcnn icon indicating copy to clipboard operation
mx-maskrcnn copied to clipboard

Published 20 hours ago • verified publisher icon TuSimple

After changeing some of the params, the speed of convergence when training RPN is amazing

Open zl1994 opened this issue 6 years ago • 0 comments

I modified the Initialization by: default And set the rpn_lr=0.002 with noing else changed. When using single GPU card, the speed of convergence is very fast. The RPNAcc can go to 0.956 at the half of the first epoch image

zl1994 avatar May 04 '18 02:05 zl1994