BEVDet
BEVDet copied to clipboard
Overlapping Bounding Boxes
Hello! Thanks for your great work! After training your model myself and visualizing the results, I found that the results often contain overlapping bounding boxes. What probably may the problem be? Is it related to erroreous dataset? Thank you very much if you can give any responses.
@huangzhengxiang The paradigm densely predicts results with different detection scores. You can filter the results according to the detection scores to remove the duplicated results