hirotomusiker / CLRerNet

The official implementation of "CLRerNet: Improving Confidence of Lane Detection with LaneIoU"
Apache License 2.0
174 stars 18 forks source link

关于车道线类别的分类 #42

Open upupbo opened 4 months ago

upupbo commented 4 months ago

请问下这边能对车道线的类别进行分类吗

hirotomusiker commented 3 months ago

Could you please rewrite your question in English? Thank you.

xiaohuo-desu commented 3 months ago

Could you please rewrite your question in English? Thank you.

Thanks for your work. Like the OP mentioned, I also wonder if it's possible to classify the different classes of lanes. I'm looking forward to identifying both solid and dashed lanes.

hirotomusiker commented 3 months ago

Thank you. In order to classify lane types, the ground-truth class ids must be provided and many changes are needed to be applied to the current code.

You need to add class bins for lane representation (currently it's just 0: no lane and 1: lane), https://github.com/hirotomusiker/CLRerNet/blob/main/libs/datasets/pipelines/lane_formatting.py#L66-L74 and modify all the codes which use the representation, for example: https://github.com/hirotomusiker/CLRerNet/blob/main/libs/models/dense_heads/clrernet_head.py#L312

So.., I have to say that it's not easy to make it possible to classify lane types with our current codebase.

xiaohuo-desu commented 3 months ago

Thank you. In order to classify lane types, the ground-truth class ids must be provided and many changes are needed to be applied to the current code.

You need to add class bins for lane representation (currently it's just 0: no lane and 1: lane), https://github.com/hirotomusiker/CLRerNet/blob/main/libs/datasets/pipelines/lane_formatting.py#L66-L74 and modify all the codes which use the representation, for example: https://github.com/hirotomusiker/CLRerNet/blob/main/libs/models/dense_heads/clrernet_head.py#L312

So.., I have to say that it's not easy to make it possible to classify lane types with our current codebase.

Thanks for your reply. Indeed, it seems not a easy work. I need to change my mind.