Reproduce “Single-Annotation, Multi-Hypothesis Segmentation with Semantic-Spatial Residual Banks” 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
Armin Dadras, Robert Sablatnig, Franziska Proksa, Markus Seidl
SSRB returns eight candidate segmentation masks trained from one visible mask per image. It combines frozen semantic predictions with image-based spatial residuals aligned to each prediction. At K=8, SSRB’s MLQ@8 on LIDC-hard is 0.8614, versus 0.8136 for Probabilistic U-Net. This measures coverage of held-out annotations, not calibrated probabilities; Probabilistic U-Net has better first-mask quality on LIDC-hard.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.