{"paper":{"title":"Expandable, Compressible, Mineable: Open-World Thermal Image Restoration","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"ECMRNet adapts to new thermal degradations by expanding isolated subspaces, pruning redundancies, and mining historical knowledge.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Huafeng Li, Jie Wen, Neng Dong, Pu Li, Wen Wang, Yafei Zhang","submitted_at":"2026-05-16T12:41:38Z","abstract_excerpt":"In open-world settings, thermal infrared (TIR) image degradations continuously emerge and evolve, while most existing all-in-one restoration methods are built on a closed-set assumption and struggle to continually adapt to novel degradations. To address this, we propose ECMRNet, an Expandable, Compressible, and Mineable Restoration Network for open-world TIR restoration from a continual learning perspective. Conceptually, ECMRNet unifies continual degradation learning as an \"expand-compress-mine\" closed-loop process, enabling sustained adaptation to new degradations with controllable evolution"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ECMRNet achieves superior overall performance across diverse single and compound degradations while using fewer parameters and lower computational cost.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that decomposing intermediate representations into group-isolated subspaces permits strict parameter isolation and fast adaptation to new degradations without interference or loss of previously learned restoration capability (stated in the structural description of ECMRNet).","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ECMRNet is a continual-learning restoration network that decomposes features into isolated groups, expands new groups for novel degradations, prunes via structural entropy, and mines historical components for compound degradations in open-world TIR imaging.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ECMRNet adapts to new thermal degradations by expanding isolated subspaces, pruning redundancies, and mining historical knowledge.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5e7f06b94235456bd923c761bbd860e28e5c94d00a6e5b173cb10a30aa1f0784"},"source":{"id":"2605.16967","kind":"arxiv","version":1},"verdict":{"id":"9e01aba0-3c2d-44c2-a1a5-7c4b5882a25c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T20:25:28.982056Z","strongest_claim":"ECMRNet achieves superior overall performance across diverse single and compound degradations while using fewer parameters and lower computational cost.","one_line_summary":"ECMRNet is a continual-learning restoration network that decomposes features into isolated groups, expands new groups for novel degradations, prunes via structural entropy, and mines historical components for compound degradations in open-world TIR imaging.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that decomposing intermediate representations into group-isolated subspaces permits strict parameter isolation and fast adaptation to new degradations without interference or loss of previously learned restoration capability (stated in the structural description of ECMRNet).","pith_extraction_headline":"ECMRNet adapts to new thermal degradations by expanding isolated subspaces, pruning redundancies, and mining historical knowledge."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16967/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T20:31:34.657075Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T20:31:19.063654Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T19:51:57.526005Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T19:50:12.031728Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.225717Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.311727Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"62e9cdbcb8a874bf1b8a1027be44f42bd70a0a16a80315031204c16b6fec68f1"},"references":{"count":76,"sample":[{"doi":"","year":null,"title":"Allrestorer: All-in-one transformer for image restoration under composite degradations","work_id":"76028e6f-f8db-4ea0-9a03-eddcea06132e","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=","work_id":"8b583880-db82-446e-a4bb-78070b2baea5","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Gustafsson and Zheng Zhao and Jens Sjölund and Thomas B","work_id":"8aec14bf-f4c2-48ee-8bcc-d72e752e022d","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Proceedings of the International Conference on Learning Representations , year=","work_id":"f8944c64-338f-4eeb-8226-5ed6350f7b63","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"arXiv preprint arXiv:2504.08219 , year=","work_id":"496945f3-9328-4f72-895f-440e565d7d29","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":76,"snapshot_sha256":"39eb457af8a2f7dd7ded52dfbd52b69b26f2b1799c4185192a77a58a5c88d3a5","internal_anchors":2},"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"}