Reproduce “ClothDojo: Learned Data Generation and Benchmarking for Bimanual Garment Flattening and Folding” 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

Zilong Huang, Zipeng Ye, Caicheng Wang, Yuchen Xie, Jiaxiang Luo

ClothDojo trains a simulator policy to generate and screen adaptive garment-flattening and folding trajectories, using human demonstrations and corrections. RGB policies trained on generated data outperform human-data policies in every tested condition for both tasks. For flattening, average success is 30.4%, versus 26.3% for human-data training. The findings are limited to the paper’s simulated tasks.

Summary by alphaXiv.

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