search

klintan / pytorch-lanenet

LaneNet implementation in PyTorch
MIT License
217 stars 60 forks source link

Instance loss 变为0 #15

Open mengpengfei opened 3 years ago

mengpengfei commented 3 years ago

Epoch 0 Epoch 1 Step 30 |(30/30)| ETA: 1.48|Total loss:1.88982|Binary loss:1.06663|Instance loss:0.82319|IoU:0.41936 97%|#########6| 29/30 [00:57<00:01, 1.96s/it]Train IoU : 0.288491758633606 Epoch 1 Epoch 2 Step 30 |(30/30)| ETA: 1.40|Total loss:0.91014|Binary loss:0.91003|Instance loss:0.00012|IoU:0.47903 97%|#########6| 29/30 [00:57<00:01, 1.96s/it]Train IoU : 0.4794849991548867 Epoch 2 Epoch 3 Step 30 |(30/30)| ETA: 1.27|Total loss:0.87450|Binary loss:0.87450|Instance loss:0.00000|IoU:0.50841 97%|#########6| 29/30 [00:57<00:01, 1.93s/it]model is saved: ./checkpoints/2_checkpoint.pth Train IoU : 0.5105022367769844 Epoch 3

从第三个epoch开始Instance loss 变为0是什么情况

IrohXu commented 3 years ago

Solve this problem, you can try my code here：https://github.com/IrohXu/lanenet-lane-detection-pytorch 解决了这个问题，你可以测试我的代码：https://github.com/IrohXu/lanenet-lane-detection-pytorch