OpenDriveLab / OpenLane-V2

[NeurIPS 2023 Track Datasets and Benchmarks] OpenLane-V2: The First Perception and Reasoning Benchmark for Road Driving
https://proceedings.neurips.cc/paper_files/paper/2023/hash/3c0a4c8c236144f1b99b7e1531debe9c-Abstract-Datasets_and_Benchmarks.html
Apache License 2.0
555 stars 66 forks source link

Rules for lane line labeling and division of lane lines for each view point #123

Open LePetitSirius opened 1 week ago

LePetitSirius commented 1 week ago

Thanks for the work on OpenLane-V2 dataset, I have the following questions looking forward to receiving your answers 1.In what order are the points labeled with the lane lines in each image? Are they labeled according to the rule from near to far? Are the labeling rules the same for each view? Are the labeling rules for lane_segment and ped_crossing the same?

  1. In each json file we get the lane lines for all views, so how do we know which view the lane belongs to? Looking forward to hearing from you. Thank you
faikit commented 1 week ago

Hi, (1) the order of points of a single lane represent the direction of the lane; (2) the lane set for a frame is for the whole BEV frame, if you want to filter out lanes for a view, maybe you can try projecting lanes onto image view and see if the projected lanes are within the image view.