Reproduce “Contrastive Bootstrap Representation Learning for Label-Efficient Agricultural Image Classification” 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

Hazem Ahmed, Ali Hamdi, Ammar Mohammed

CBRL combines contrastive learning, which separates different images, with bootstrap prediction of an exponential-moving-average target representation, then fine-tunes on scarce labels. On DeepWeeds, using 10% of labels and one seed, it reaches 79.77% test macro-F1, versus 47.49% for random initialization and 63.99% for denoising. These are single-run point estimates, not evidence of statistical superiority.

Summary by alphaXiv.

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