search

MaybeShewill-CV / lanenet-lane-detection

Unofficial implemention of lanenet model for real time lane detection
Apache License 2.0
2.36k stars 886 forks source link

Need help with training for two classes with custom dataset #519

Closed chee0007 closed 3 years ago

chee0007 commented 3 years ago

Hi, I am trying to train two classes - Lane and Curbs using bisenetv2 I labelled lanes as 1, curbs as 2 in the binary_seg mask I have set NUM_CLASSES: 2, FRONT_END: 'bisenetv2' in tusimple_lanenet.yaml

The training was successful. But when I run test_lanenet.py, the binary_seg_image output is black without any labels, and the instance_seg_image looks normal with different coloured lane clusters.

The binary_seg_image output should have labels 1 and 2, but I am not getting any. Please advise if I have followed the procedures correctly. Thanks for your help.

MaybeShewill-CV commented 3 years ago

@chee0007 The postprocess code here was designed for tusimple dataset. You need to adjust those code yourself or you could simple add binary seg mask upon origin image:)

chee0007 commented 3 years ago

@MaybeShewill-CV Thanks! I will try it. :)

Petros626 commented 8 months ago

@chee0007 you were successfull?