Reproduce “Learning to Discover Novel Visual Categories via Deep Transfer Clustering” 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

Kai Han, Andrea Vedaldi, Andrew Zisserman

The authors introduce deep transfer clustering (DTC) to group unlabeled images into new categories, using labeled examples from different categories to initialize a feature representation that is refined alongside cluster assignments. A probe-based component estimates the number of clusters. On known-five-class CIFAR-10, DTC-Π reaches 87.5% clustering accuracy versus MCL’s 64.2%, averaged over 10 runs; specific to that benchmark.

Summary by alphaXiv.

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