About CULane official evaluation function

[huanFigure]

Hello, I have run the enet_baseline model,and I have gotten the result 'checkpoints/enetbaseline/culane/model.pt'.And then I the test enetmodel ,and got the output 'output/driverframe/.MP4/*.lines.txt'.But when I run './tools/culane_evaluation/eval.sh' to evaluate the output files, I got the number 0 or -1

Have you ever encountered such a situation? And the configuration in eval.sh: data_dir is the original CULane data which download from website detect_dir is the output of the test 'python main_landet.py --test --config=configs/lane_detection/baseline/enet_culane.py --mixed-precision' Is there anything wrong with my configuation?

[huanFigure]

And the eval class configuration 'output/out*.txt' ,I copy the downloaded from website CULane to this directory and renamed them

[voldemortX]

@huanFigure I only encounter this when the model really have 0 accuracy. Did you train the model yourself? Perhaps try testing one of the uploaded checkpoints and verify if the issue is in evaluation script?

Also could you check if the predicted lines.txt have normal contents?

[huanFigure]

Thanks for your replay. I have solved this problem. The reason is that the file name generated by the test end with '.jpg.lines.txt' which doesn't match the original file name '.lines.txt'