Reproduce “Active Learning under Simulated Class Imbalance and Label Shift in Binary Semantic Segmentation” 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
Julien Combes, Jean-François Coeurjolly, Alexandre Derville
The authors conduct a controlled study selecting images for binary defect segmentation, independently varying pool and test defect prevalence. Entropy and core-set selection often outperform random selection when pool defects are scarce; on patched potato diseases, entropy gains +0.344 F1 units. The advantage is negligible at pool defect rates of 95% or more, and mismatch can degrade all strategies.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.