Reproduce “Attend, Infer, Repeat: Fast Scene Understanding with Generative Models” 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
S. M. Ali Eslami, Nicolas Heess, Theophane Weber, Yuval Tassa, David Szepesvari, Koray Kavukcuoglu, Geoffrey E. Hinton
AIR infers scene descriptions by recurrently attending to image regions, estimating object identity and pose, and stopping when no objects remain. On synthetic digit scenes containing at most two objects, AIR decomposes images; DAIR generalizes better when tested on changed counts, though standard AIR fails on three-digit extrapolation. Three inference steps averaged 17 ms per image, and real-table transfer remains challenging.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.