search

lucastabelini / PolyLaneNet

Code for the paper entitled "PolyLaneNet: Lane Estimation via Deep Polynomial Regression" (ICPR 2020)
https://arxiv.org/abs/2004.10924
MIT License
295 stars 74 forks source link

why my output is like this #10

Closed feifeiwei closed 4 years ago

feifeiwei commented 4 years ago

Hello,thanks for your great work, when i train polyLaneNet with my dataset about 30000 images within about 50 epochs i run the demo like this, what's the problem with it？

lucastabelini commented 4 years ago

How exactly did you train with your dataset? There may be a bug in your dataset loader.

feifeiwei commented 4 years ago

my dataloader seem to be right, How much does your poly loss finally converge to? my poly loss is from about 100 to 1~3 within 50 epochs, The question is whether poly loss doesn't converge？

lucastabelini commented 4 years ago

On TuSimple the loss converges to ~0.05. You can't just look at the loss value though. If the annotations are not being loaded correctly the model might still converge, but not to what you want. Try visualizing your data after it has been loaded by LaneDataset. You can do this by running python -m lib.datasets.lane_dataset, which will load the dataset described in the file config.yaml.

Since in every image you showed the predictions look the same, it seems the model is converging to something weird, which is generally caused by bad data.

feifeiwei commented 4 years ago

There seems to be wrong with my dataset.py script, thanks for your reply.