In segmentation_baselines directories, the baseline models (HDMapNet, FusionNet) are shared. However, it seems that the results are only for single-frame, and the multi-frame results are not available?
Specifically, I mean the frame merging code mentioned as below:
"To adapt it to our multi-frame detection task, we apply the frame merging method proposed in [25] to construct the world-level graph of lane centerlines."
hi,
Thanks for sharing your code,I read the readme in folder “segmentation_baselines”,and found there is no train.py to train segmentation_baselines,Look forward to your favourable reply。
Thanks for your excellent work!
I want to know how you got the gt centerlines of the maps. Did you calculate them by the boundaries or use other auxiliary tools?
Thanks again for your reply!
Hello, thanks for your interesting work! I just have a question about the behavior cloning used in the paper? It seems that this code-base doesn't really use the technical about imitation learning, is there anything wrong with me? Very looking forward to your reply and de-confusion.