Supplementary Material: Progressive Coordinate Transforms for Monocular 3D Object Detection
–Neural Information Processing Systems
In this supplementary material, we provide additional experimental results and qualitative visualizations. Specifically, we demonstrate the impacts of using different off-the-shelf models in Sec. 2, We show that our proposed PCT method achieves consistent improvements with all configurations. Sec. 3. We present both successful predictions and failure cases. In this section, we firstly show the additional results on "Pedestrian" and "Cyclist" categories.Then, At last, we would like to demonstrate the impacts of using different 2D detectors and depth estimators on the performance of 3D detection, which is the first two steps of the coordinate-based methods as mentioned in Sec. 3 of the main submission. Non-rigid structures and various shape make it more challenging for monocular 3D detection to accurately detect "Pedestrian" and "Cyclist".
Neural Information Processing Systems
Nov-14-2025, 11:04:01 GMT
- Country:
- South America > Brazil (0.05)
- Technology:
- Information Technology > Artificial Intelligence > Vision (0.69)