Reproduce “GraCo: Granularity-Controllable Interactive 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
Yian Zhao, Kehan Li, Zesen Cheng, Pengchong Qiao, Xiawu Zheng, Rongrong Ji, Chang Liu, Li Yuan and 1 more
This paper presents GraCo, a method for granularity-controllable interactive segmentation that allows users to precisely dictate the desired segmentation detail. GraCo introduces an automated data generation pipeline that quantifies mask granularity and leverages it to train a model capable of producing segmentations at specified levels of detail, demonstrating improved efficiency and flexibility compared to existing methods.
Key insights
- Automated generation of diverse, multi-granularity mask-granularity pairs (AGG) is feasible and highly effective, circumventing the need for expensive manual annotation.
- Introducing an explicit granularity control parameter allows for precise, user-defined segmentation output, resolving spatial ambiguity without generating redundant masks.
- Granularity can be robustly quantified by combining scale and semantic attributes, enabling a model to learn and produce segmentations consistent with human intuition.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.