Reproduce “Unsupervised learning of features and object boundaries from local prediction” 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
Heiko H. Schütt, Wei Ji Ma
The authors introduce an unsupervised vision model that jointly learns image features and local connections by predicting relationships between nearby locations. The inferred connections indicate which locations may belong together and support contour extraction through additional graph processing. On BSDS500, first-layer Predseg1 reaches F(ODS) 0.74, versus 0.73 for gPb-owt-ucm; deeper Predseg1 performs substantially worse, and supervised contour systems score higher.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.