search

MaybeShewill-CV / lanenet-lane-detection

Unofficial implemention of lanenet model for real time lane detection
Apache License 2.0
2.38k stars 888 forks source link

I am getting Val loss : nan. Help me out please! #547

Closed Berrlinn closed 2 years ago

Berrlinn commented 2 years ago

2022-03-29 20:59:05.616 | INFO | trainner.tusimple_lanenet_multi_gpu_trainner:train:531 - => Epoch: 98 Time: 2022-03-29 20:59:05 Train loss: 5.46163 Train miou: 0.36632 Val loss: nan Val miou: nan... train loss: 5.47592, b_loss: 1.01594, i_loss: 0.90328, miou: 0.36703: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████| 1/1 [00:05<00:00, 5.19s/it] 0it [00:00, ?it/s] 2022-03-29 20:59:10.806 | INFO | trainner.tusimple_lanenet_multi_gpu_trainner:train:531 - => Epoch: 99 Time: 2022-03-29 20:59:10 Train loss: 5.47592 Train miou: 0.36703 Val loss: nan Val miou: nan... 2022-03-29 20:59:10.806 | INFO | trainner.tusimple_lanenet_multi_gpu_trainner:train:543 - Best model's val mious are: [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan] 2022-03-29 20:59:10.806 | INFO | trainner.tusimple_lanenet_multi_gpu_trainner:train:544 - Complete training process good luck!!

MaybeShewill-CV commented 2 years ago

@Berrlinn Like I said check your label data first:)

pankaj251003 commented 2 years ago

@Berrlinn Have you solved the Val loss Nan problem .....if yes please help

Berrlinn commented 2 years ago

@Berrlinn Have you solved the Val loss Nan problem .....if yes please help

No, I tried it for a long time but was unable to resolve the error. So, I had to leave it. Did you found any solution?

Berrlinn commented 2 years ago

@Berrlinn Have you solved the Val loss Nan problem .....if yes please help

But you can try this repo "https://github.com/harryhan618/LaneNet". I was able to complete the training process from this repo.