Reproduce “A Benchmark for Spatially Grounded Gesture Generation” 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
Anna Deichler, Rishabh Dabral, Fethiye Irmak Dogan, Anindita Ghosh, Jonas Beskow
The authors introduce a benchmark for evaluating referential gesture generation with known 3D targets. Systems receive target coordinates, so the benchmark tests gesture timing and forearm aim, not referent resolution. On a 100-item held-out MM-Conv set, RePointer scores 0.905 spatially, versus 0.767 for captured human motion. Its naturalness rating was not significantly higher than the other generated system, and listener identification was not tested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.