PytorchSSD icon indicating copy to clipboard operation
PytorchSSD copied to clipboard

RefineDet320

Open shwoo1219 opened this issue 7 years ago • 4 comments

Hello thank you for sharing your codes

There is one issue in RefineDet320 training. The training stops after training epoch of 10. Even I tried several times, always the same phenomenon occurs. Any suggestion?

Thanks

shwoo1219 avatar Mar 30 '18 14:03 shwoo1219

I have the same problem and try to solve it now. The dataloader always stops after the testing phase. Only testing the dataset at last can solve this problem temporarily.

lzx1413 avatar Mar 30 '18 14:03 lzx1413

Thanks for your reply.

Another issue in your RefineDet320 code. The transfer connection block is implemented as following. x = F.relu(F.relu(l(u(x)),inplace=True)+t,inplace=True)

However, according to the paper, it should be as following. x = F.relu(l(F.relu(u(x)+t,inplace=True)),inplace=True)

shwoo1219 avatar Mar 31 '18 15:03 shwoo1219

Thanks, I have not reproduced the result mentioned in the original paper. There must be some wrong with my code.

lzx1413 avatar Apr 01 '18 02:04 lzx1413

The test stage only supports one gpu.

ghost avatar Mar 11 '19 01:03 ghost