{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ITLKNZFONYJVUG6BFIRJWJOI5I","short_pith_number":"pith:ITLKNZFO","schema_version":"1.0","canonical_sha256":"44d6a6e4ae6e135a1bc12a229b25c8ea0c54ae4b9269dcc373d8d8d325c201d0","source":{"kind":"arxiv","id":"2502.05540","version":3},"attestation_state":"computed","paper":{"title":"Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"De Cheng, Di Xu, Peng Wang, Qirui Wu, Shizhou Zhang, Yanning Zhang, Yinghui Xing","submitted_at":"2025-02-08T12:10:02Z","abstract_excerpt":"Catastrophic forgetting is a critical chanllenge for incremental object detection (IOD). Most existing methods treat the detector monolithically, relying on instance replay or knowledge distillation without analyzing component-specific forgetting. Through dissection of Faster R-CNN, we reveal a key insight: Catastrophic forgetting is predominantly localized to the RoI Head classifier, while regressors retain robustness across incremental stages. This finding challenges conventional assumptions, motivating us to develop a framework termed NSGP-RePRE. Regional Prototype Replay (RePRE) mitigates "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2502.05540","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-02-08T12:10:02Z","cross_cats_sorted":[],"title_canon_sha256":"006f8016cd1d80aa999336795cba263f60d3582ad3ff036cd95236956c603643","abstract_canon_sha256":"0e2253ccbc5eebd18b895b588e96ead7c9c4a52f2e80982b5669039c5c7e1be0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:41.060975Z","signature_b64":"bDXY8cv7hwGBjLUbfR40IW1inufMkaFxn6FfaK2LutNqCuL22MizUbtoXkTkX0n7jWVEQaAurxesWx9mHISCBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"44d6a6e4ae6e135a1bc12a229b25c8ea0c54ae4b9269dcc373d8d8d325c201d0","last_reissued_at":"2026-07-05T11:11:41.060495Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:41.060495Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"De Cheng, Di Xu, Peng Wang, Qirui Wu, Shizhou Zhang, Yanning Zhang, Yinghui Xing","submitted_at":"2025-02-08T12:10:02Z","abstract_excerpt":"Catastrophic forgetting is a critical chanllenge for incremental object detection (IOD). Most existing methods treat the detector monolithically, relying on instance replay or knowledge distillation without analyzing component-specific forgetting. Through dissection of Faster R-CNN, we reveal a key insight: Catastrophic forgetting is predominantly localized to the RoI Head classifier, while regressors retain robustness across incremental stages. This finding challenges conventional assumptions, motivating us to develop a framework termed NSGP-RePRE. Regional Prototype Replay (RePRE) mitigates "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.05540","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2502.05540/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2502.05540","created_at":"2026-07-05T11:11:41.060547+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.05540v3","created_at":"2026-07-05T11:11:41.060547+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.05540","created_at":"2026-07-05T11:11:41.060547+00:00"},{"alias_kind":"pith_short_12","alias_value":"ITLKNZFONYJV","created_at":"2026-07-05T11:11:41.060547+00:00"},{"alias_kind":"pith_short_16","alias_value":"ITLKNZFONYJVUG6B","created_at":"2026-07-05T11:11:41.060547+00:00"},{"alias_kind":"pith_short_8","alias_value":"ITLKNZFO","created_at":"2026-07-05T11:11:41.060547+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06978","citing_title":"CL-CLIP: CLIP-Based Continual Learning Framework with Cost-Volume Category Decoupling for Object Detection","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20738","citing_title":"STAR-IOD: Scale-decoupled Topology Alignment with Pseudo-label Refinement for Remote Sensing Incremental Object Detection","ref_index":167,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I","json":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I.json","graph_json":"https://pith.science/api/pith-number/ITLKNZFONYJVUG6BFIRJWJOI5I/graph.json","events_json":"https://pith.science/api/pith-number/ITLKNZFONYJVUG6BFIRJWJOI5I/events.json","paper":"https://pith.science/paper/ITLKNZFO"},"agent_actions":{"view_html":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I","download_json":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I.json","view_paper":"https://pith.science/paper/ITLKNZFO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.05540&json=true","fetch_graph":"https://pith.science/api/pith-number/ITLKNZFONYJVUG6BFIRJWJOI5I/graph.json","fetch_events":"https://pith.science/api/pith-number/ITLKNZFONYJVUG6BFIRJWJOI5I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I/action/storage_attestation","attest_author":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I/action/author_attestation","sign_citation":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I/action/citation_signature","submit_replication":"https://pith.science/pith/ITLKNZFONYJVUG6BFIRJWJOI5I/action/replication_record"}},"created_at":"2026-07-05T11:11:41.060547+00:00","updated_at":"2026-07-05T11:11:41.060547+00:00"}