simple-faster-rcnn-pytorch
simple-faster-rcnn-pytorch copied to clipboard
About mAP
Thanks for your great work!
But I met some troubles in training, the mAP stayed the same after 0 epoch and wouldn't change any more!
I print the mAP every epoch, it outputs as bellow:
model construct completed 2829it [54:19, 1.15s/it] 2993it [23:02, 2.16it/s]
epoch 0 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:50, 2.18it/s]
epoch 1 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:59, 2.17it/s]
epoch 2 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:54, 2.18it/s]
epoch 3 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:44, 2.19it/s]
epoch 4 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:50, 2.18it/s]
epoch 5 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:50, 2.18it/s]
epoch 6 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:47, 2.19it/s]
epoch 7 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:43, 2.20it/s]
epoch 8 mAP is 0.3790 ...
0it [00:00, ?it/s] 2993it [22:48, 2.19it/s]
epoch 9 mAP is 0.3790 ...
hi, where is the map code in this package,i find noting here!
I have the same trouble that the mAP is about 0.3 when I retrain the model with the epoches of 100.