Reproduce “Temporal Recurrent Networks for Online Action Detection” 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

Mingze Xu, Mingfei Gao, Yi-Ting Chen, Larry S. Davis, David J. Crandall

TRN forecasts upcoming actions and feeds that predicted context into a present-time online action decision using historical observations. Its decoder predicts future action scores, and a gate combines predicted hidden states with current visual features and state. On TVSeries, TRN reaches 83.7% mcAP versus RED’s 79.2% with two-stream features; the benchmark comparison does not measure runtime or driving outcomes.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.