Giters

cfzd / Ultra-Fast-Lane-Detection

Ultra Fast Structure-aware Deep Lane Detection (ECCV 2020)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Where do you predict category on Culane dataset?

zillur-av opened this issue · comments

commented

I see you can predict on lane location using the following line and some post processing. But in Culane dataset, beside location, you predict on 9 categories too. Where is the network output code for that.?

Ultra-Fast-Lane-Detection/model/model.py

Line 90 in 353df10

group_cls = self.cls(fea).view(-1, *self.cls_dim)

commented

@zillur-av
We don't predict the category. CULane provides the list files of each category, so we can test on each category.

commented

Ok. So, if I understand correctly, you tack an image, let's say night category, and if the model can detect lanes on that image, you say it is detecting night category, something like that.

commented

@zillur-av
Yes. CULane provides a list file containing all the images in the night category. We read all the images in the list, and eval on these images. The evaled results are the result of this category. It is the same for other categories like shadow, dazzle, etc.