Reproduce “Resource-Efficient Affordance Grounding with Complementary Depth and Semantic Prompts” 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
Yizhou Huang, Fan Yang, Guoliang Zhu, Gen Li, Hao Shi, Yukun Zuo, Wenrui Chen, Zhiyong Li and 1 more
BiT-Align predicts activation maps for requested actions on inactive objects using human–object interaction images and action labels for weak supervision. It injects pseudo-depth cues into an RGB encoder and uses action text to guide visual attention. On AGD20K Seen, it achieves KLD of 1.105 versus WSMA’s 1.176, while using fewer trainable parameters; results are activation-map comparisons, not pixel-accurate segmentation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.