search

cardwing / Codes-for-Lane-Detection

Learning Lightweight Lane Detection CNNs by Self Attention Distillation (ICCV 2019)
MIT License
1.04k stars 333 forks source link

ERFNet pretrained model #140

Closed jcdubron closed 5 years ago

jcdubron commented 5 years ago

Thanks for your codes for lane detection. I wonder where the ERFNet pretrained model comes from or could you provide more details about pretraining setting?

In train_erfnet.sh, the image shape is (208, 976) that does not correspond to the ratio of (590, 1640). Why?

cardwing commented 5 years ago

I just use the original ERFNet model as the pretrained model, which is trained on Cityscape dataset. You can also train the model from scratch. As to the image size, I crop the original image to remove the upper part which is totally background and then resize the cropped image to (208, 976).

jcdubron commented 5 years ago

Thank you for your reply.