{"paper":{"title":"Sequential Feature Selection for Efficient Landslide Segmentation from Multi-Spectral Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Sequential forward floating selection finds an 8-channel subset that matches or exceeds full 30-channel accuracy for landslide segmentation.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Arsalaan Ahmad, Oktay Karakus, Paul L. Rosin","submitted_at":"2026-05-10T20:46:30Z","abstract_excerpt":"Landslide detection from satellite imagery has advanced through deep learning, yet most models rely on large, highly correlated spectral-topographic inputs whose contributions remain poorly understood. The question of which channels are actually necessary has received surprisingly little attention. This matters: redundant or correlated inputs obscure physical interpretability, inflate computational overhead, and can actively degrade model performance through the Hughes Phenomenon. We present a systematic, explainable channel-selection framework for the Landslide4Sense benchmark, combining Sent"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Beyond identifying a compact 8-channel subset that matches or exceeds the segmentation F1 of configurations using up to 30 channels, we use the selection process itself to interrogate which spectral and topographic features landslide models genuinely rely on, and what this reveals about the physical cues driving their predictions.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the lightweight U-Net++ proxy model used during SFFS iterations accurately captures interaction effects and performance trends of the final full-scale segmentation model, and that single-band drop tests miss interactions while SFFS does not.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Sequential Forward Floating Selection with a U-Net++ proxy identifies an 8-channel subset from multi-spectral and terrain data that matches or exceeds F1 scores of full 30-channel configurations for landslide segmentation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Sequential forward floating selection finds an 8-channel subset that matches or exceeds full 30-channel accuracy for landslide segmentation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"87ec8638075d1c63f65741cf8c40d62a790ad487b804b97645c542a1813cd735"},"source":{"id":"2605.09746","kind":"arxiv","version":2},"verdict":{"id":"51eaf8a4-0097-4011-a407-680592e7ca0d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T02:36:46.853842Z","strongest_claim":"Beyond identifying a compact 8-channel subset that matches or exceeds the segmentation F1 of configurations using up to 30 channels, we use the selection process itself to interrogate which spectral and topographic features landslide models genuinely rely on, and what this reveals about the physical cues driving their predictions.","one_line_summary":"Sequential Forward Floating Selection with a U-Net++ proxy identifies an 8-channel subset from multi-spectral and terrain data that matches or exceeds F1 scores of full 30-channel configurations for landslide segmentation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the lightweight U-Net++ proxy model used during SFFS iterations accurately captures interaction effects and performance trends of the final full-scale segmentation model, and that single-band drop tests miss interactions while SFFS does not.","pith_extraction_headline":"Sequential forward floating selection finds an 8-channel subset that matches or exceeds full 30-channel accuracy for landslide segmentation."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.09746/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-20T07:02:01.438347Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T16:36:55.768135Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T12:31:18.050570Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T09:58:56.840035Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"355199cc1b532e5f9e568e81d04582b4136af726002d9c7cb32f1dd4f699b02e"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}