Reproduce “Fine-Grained Image-Text Correspondence with Cost Aggregation for Open-Vocabulary Part 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
Jiho Choi, Seonho Lee, Minhyun Lee, Seungho Lee, Hyunjung Shim
PartCATSeg performs zero-shot open-vocabulary part segmentation by matching objects and generic parts separately, then combining them to score object-specific part labels. It adds a compositional loss aligning part and object predictions, and uses DINO features for spatial guidance. On Pascal-Part-116 Pred-All, it achieves 45.77 h-IoU versus 30.67 for PartCLIPSeg; Pred-All requires predictions without object masks or labels.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.