Reproduce “Self-Supervised Drivable Area and Road Anomaly Segmentation using RGB-D Data for Robotic Wheelchairs” 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
Hengli Wang, Yuxiang Sun, Ming Liu
The authors propose a self-supervised method that uses RGB-D images to generate noisy pixel labels for wheelchair-drivable ground, road anomalies, and unknown regions, then trains segmentation networks on them. On a held-out set of 585 manually labeled images, RTFNet trained on these labels achieves 72.26% All mean IoU, versus 62.20% for Mix-and-Match. The result is limited to this dataset and evaluation protocol; manually trained networks perform better.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.