search

wvangansbeke / LaneDetection_End2End

End-to-end Lane Detection for Self-Driving Cars (ICCV 2019 Workshop)
https://arxiv.org/pdf/1902.00293.pdf
Other
643 stars 216 forks source link

python main.py #53

Closed xuxiuzhi2627 closed 4 years ago

xuxiuzhi2627 commented 5 years ago

python main.py --image_dir ./Labels/images/ --gt_dir ./Labels/ground_truth/ --end_to_end True size train loader is 2901 size valid loader is 725 Traceback (most recent call last): File "main.py", line 548, in main() File "main.py", line 89, in main split_percentage=args.split_percentage) File "/home/chengpeisong/Downloads/LaneDetection_End2End-master/Backprojection_Loss/Dataloader/Load_Data_new.py", line 277, in get_loader nclasses=nclasses) File "/home/chengpeisong/Downloads/LaneDetection_End2End-master/Backprojection_Loss/Dataloader/Load_Data_new.py", line 92, in init assert len(self.rgb_lst) == len(self.gt_lst) == 3626 AssertionError

How to solve it ?The image_dir is the folder images of dir, right?

wvangansbeke commented 5 years ago

Hi @xxzcpsmemeda,

The code for the backprojection loss works on the complete TuSimple dataset (=3626 images). The link is in the README.

Best, Wouter

oesni commented 5 years ago

So, Do I need to create ground truth png images for complete TuSimple dataset(=3626 images)?? Your link contains only 2535 jpg images and 2535 ground truth png images.

wvangansbeke commented 5 years ago

Hi @oesni,

Yes. For the conventional semantic segmentation approach you'll need the ground truth maps. It's very easy to make them though. I only included the ground truth for the egolane experiments.

Best, Wouter