Reproduce “3DRef: 3D Dataset and Benchmark for Reflection Detection in RGB and Lidar Data” 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
Xiting Zhao, Sören Schwertfeger
3DRef provides a labeled 3D testbed for evaluating reflection segmentation in LiDAR point clouds and RGB images. It uses a textured, labeled mesh and ray-casting to generate annotations for data from three LiDAR sensors and RGB imagery. The dataset contains 48,024 labeled point clouds and 3,799 labeled images. On 3DRef-Glass, EBLNet glass mIoU rises from 60.49 to 86.71 after retraining, a task-specific result, not evidence of deployment or broad generalization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.