Reproduce “INTERACT: Learning Interaction-Aware World Dynamics from Driving Videos for Autonomous Driving” 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
Zhan Chen, Songyan Zhang, Wenhui Huang, Zirui Li, Zhidong Wang, Qihang Huang, Shiyu Fang, Chen Lv
INTERACT learns persistent scene slots from driving video, infers latent ego actions, and predicts action-conditioned future slots to train a trajectory planner. Its first two stages use self-supervision without object annotations, while the planner uses historical inputs at inference. On NAVSIM-v1 navtest, it scores 93.8 PDMS versus 93.7 for DrivoR, in a four-second non-reactive simulation, not real-world crash reduction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.