Reproduce “Weakly Supervised Monocular 3D Object Detection using Multi-View Projection and Direction Consistency” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Runzhou Tao, Wencheng Han, Zhongying Qiu, Cheng-zhong Xu, Jianbing Shen
WeakMono3D trains monocular 3D object detection from image-level 2D boxes and direction labels, using paired views and geometric consistency to constrain predictions. At inference, it uses one image. On the KITTI car test set, at IoU 0.7 and Moderate difficulty, it reports 8.80 APBEV and 4.85 AP3D, versus WeakM3D’s 5.66 and 2.26; however, WeakM3D uses LiDAR during training, while WeakMono3D does not.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.