Reproduce “Learning Language-Conditioned Traversability Representations for Adaptive Visual Navigation” 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
Senda Chen, Changxu Cheng, Fangdi Li, Tao Wang, Wuyue Zhao
LaTraNav uses a vision-language model to represent traversable regions for a path planner. The model predicts a goal and learned mask-token states, which the planner combines with current visual features to generate paths; training uses both mask and path supervision. On 11,768 feasible tasks, it achieves 63.95% Joint_all versus 59.83% with an explicit-mask planner, in a multi-source test set.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.