Reproduce “Noisy Annotations in 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
Moshe Kimhi, Omer Kerem, Eden Grad, Ehud Rivlin, Chaim Baskin
The authors build benchmarks that test instance segmentation under five kinds of annotation error, including distorted masks and missing objects. They evaluate standard-trained Mask R-CNN with a ResNet-50 on COCO-N, where mask mAP falls from 34.6 on clean annotations to 22.3 on Hard. This shows vulnerability to the tested corruptions, not their prevalence in ordinary datasets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.