Open zaie opened 5 years ago
Are you working on nuScenes or KITTI datasets? According to the KITTI benchmark, pointpillars can obtain normal cyclist results. However, in nuScenes dataset, I get 0 mAP on cyclist class.
Me too..
Can I ask what performance do you get on the pedestrian class of the Nuscenes? Cannot reproduce the results of 60+.
Thanks for this wonderful implementation. I noticed that the mAP of bicycle is 0, which seems abnormal. Is there any potential bugs resulting in such performance or Point Pillars is not suitable for bicycle detection? Thanks for your help.