ETSAM adapts Segment Anything 2 to treat 3D scans as video, scoring 0.812 on membrane segmentation
bravo_abad · x · 2026-10-21
Selvaraj and Cheng adapted Segment Anything Model 2 into ETSAM for membrane segmentation in electron microscopy: successive slices of a 3D scan are treated as video frames, letting SAM2's memory carry membrane information across slices where a single slice may show the structure faintly. The method was trained on 83 experimental and 28 simulated scans with a second-stage refinement; the unmodified model failed entirely with automated prompts. Across 17 test scans ETSAM scored 0.812, beating baselines, with visual comparisons against Membrain-Seg and TARDIS.
More from Research
- Explore Qwen3-32B neurons layer by layer on Neuronpedia — burny_tech · 2026-10-21
- The Core Tension of Formal Mechanistic Interpretability: Fuzzy Concepts Defy Symbolic Verification — burny_tech · 2026-10-21
- Formal verification is only as good as your specification: the Lean caveat — burny_tech · 2026-10-21
- Dream-1 controlled 31 unseen robot types across 53 countries within 24 hours — gharik · 2026-10-21
- ICLR 2026 paper brings provable guarantees to automated circuit discovery in mechanistic interpretability — burny_tech · 2026-10-21
- Model Nails Ligand Position on Homolog with Known Structure — johnowhitaker · 2026-10-21