AI Paper Atlas

Research map / 3D vision / 3D generation

SeisMark: A Large-Scale Open Benchmark for Robust 3D Seismic Fault Detection

Min Jun Park, Joseph Stitt, Robert Graham Clapp, Ilan Naiman, Artem Goncharuk, Kevin F. Smith

ICML 2026 · 3D vision › 3D generation · approach: Learning 3D reconstruction from images

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
evaluate on high-intensity styled texture instead of CleanFaultSeg3D (V1) · SeisMark, recall at T=1recall: 0.804 -> 0.305 (Aus-High)Hurt
train with real-data labeled slices (Fault-Net) versus synthetic-only detectorsFault-Net · SeisMark styled variantsIoU: 0.384 vs 0.237 (RGF, Australia Low)Improved
curriculum of more realistic noise (V2 vs V1)FaultSeg3D+ · SeisMark styled variantsrecall at T=1: 0.305 -> 0.388 (Aus-High)Improved
diffusion styling at low intensity versus CleanFaultSeg3D+ · SeisMark, precision at T=1precision: 0.676 -> 0.902 (F3-Low)Improved
diffusion styling at low intensity versus CleanFault-Net · SeisMark, precision at T=1precision: 0.805 -> 0.868 (F3-Low)Improved
raise texture intensity from Low to HighNorth Sea style · structural fidelitySSIM: 0.92 -> 0.88Hurt
raise texture intensity from Low to HighAustralia style · structural fidelityGMC: 0.93 -> 0.90Hurt
raise texture intensity from Low to HighRGF · SeisMarkIoU: 0.238 -> 0.151 (North Sea)Hurt
evaluate on Clean data onlyFaultSeg3D, FaultSeg3D+, Fault-Net · SeisMark Cleanrecall at T=1: Recall ≈0.79–0.81No clear effect
strict tolerance T=0 versus T>=1all detectors · SeisMarkF1: 0.415 -> 0.549 (V1 Clean, T=0 to T=1)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy