{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YJCNZ7CTLVHNWRFVSLYTGLEH5U","short_pith_number":"pith:YJCNZ7CT","schema_version":"1.0","canonical_sha256":"c244dcfc535d4edb44b592f1332c87ed28ff274abd3112d3a8775f67e1ac1648","source":{"kind":"arxiv","id":"2512.08227","version":1},"attestation_state":"computed","paper":{"title":"New VVC profiles targeting Feature Coding for Machines","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Three simplified VVC profiles deliver up to 95% faster encoding of neural network features for machines with minimal rate penalty.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ashan Perera, Hari Kalva, Juan Merlos, Md Eimran Hossain Eimon, Velibor Adzic","submitted_at":"2025-12-09T04:13:07Z","abstract_excerpt":"Modern video codecs have been extensively optimized to preserve perceptual quality, leveraging models of the human visual system. However, in split inference systems-where intermediate features from neural network are transmitted instead of pixel data-these assumptions no longer apply. Intermediate features are abstract, sparse, and task-specific, making perceptual fidelity irrelevant. In this paper, we investigate the use of Versatile Video Coding (VVC) for compressing such features under the MPEG-AI Feature Coding for Machines (FCM) standard. We perform a tool-level analysis to understand th"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2512.08227","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CV","submitted_at":"2025-12-09T04:13:07Z","cross_cats_sorted":[],"title_canon_sha256":"ccfc3ff335ad8bf383f1c58f490861311398826ecdc4a9522f7ac6702be7e2c4","abstract_canon_sha256":"26b0236ce24a3e57a9e94e9fd121af0bba703c3ea8c2db6a4d026955a6496f73"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-15T00:21:15.234902Z","signature_b64":"pXO2fZ8vbDi6A0ovVXEp1LtpvMqOk0aScZ7YaSTxKFJIQdarfBuxv6i0K5EvcZZbOkSlyq6EpoTPXVo1I636AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c244dcfc535d4edb44b592f1332c87ed28ff274abd3112d3a8775f67e1ac1648","last_reissued_at":"2026-07-15T00:21:15.233972Z","signature_status":"signed_v1","first_computed_at":"2026-07-15T00:21:15.233972Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"New VVC profiles targeting Feature Coding for Machines","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Three simplified VVC profiles deliver up to 95% faster encoding of neural network features for machines with minimal rate penalty.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ashan Perera, Hari Kalva, Juan Merlos, Md Eimran Hossain Eimon, Velibor Adzic","submitted_at":"2025-12-09T04:13:07Z","abstract_excerpt":"Modern video codecs have been extensively optimized to preserve perceptual quality, leveraging models of the human visual system. However, in split inference systems-where intermediate features from neural network are transmitted instead of pixel data-these assumptions no longer apply. Intermediate features are abstract, sparse, and task-specific, making perceptual fidelity irrelevant. In this paper, we investigate the use of Versatile Video Coding (VVC) for compressing such features under the MPEG-AI Feature Coding for Machines (FCM) standard. We perform a tool-level analysis to understand th"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose three lightweight essential VVC profiles-Fast, Faster, and Fastest. The Fast profile provides 2.96% BD-Rate gain while reducing encoding time by 21.8%. Faster achieves a 1.85% BD-Rate gain with a 51.5% speedup. Fastest reduces encoding time by 95.6% with only a 1.71% loss in BD-Rate.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the tool-level impact observed on the tested features and tasks will hold for the full range of FCM use cases and that the proposed profile simplifications do not introduce unmeasured accuracy drops on unseen models or datasets.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Three lightweight VVC profiles for feature coding achieve up to 2.96% BD-Rate gain and 95.6% encoding speedup while preserving downstream task accuracy under the MPEG-AI FCM framework.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Three simplified VVC profiles deliver up to 95% faster encoding of neural network features for machines with minimal rate penalty.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"4d8275460902a0e52b25f35387366df7975321275e2216a66459def0fee57fff"},"source":{"id":"2512.08227","kind":"arxiv","version":1},"verdict":{"id":"df6697f1-18b7-4898-924a-b4c580887b98","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T00:22:33.599277Z","strongest_claim":"We propose three lightweight essential VVC profiles-Fast, Faster, and Fastest. The Fast profile provides 2.96% BD-Rate gain while reducing encoding time by 21.8%. Faster achieves a 1.85% BD-Rate gain with a 51.5% speedup. Fastest reduces encoding time by 95.6% with only a 1.71% loss in BD-Rate.","one_line_summary":"Three lightweight VVC profiles for feature coding achieve up to 2.96% BD-Rate gain and 95.6% encoding speedup while preserving downstream task accuracy under the MPEG-AI FCM framework.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the tool-level impact observed on the tested features and tasks will hold for the full range of FCM use cases and that the proposed profile simplifications do not introduce unmeasured accuracy drops on unseen models or datasets.","pith_extraction_headline":"Three simplified VVC profiles deliver up to 95% faster encoding of neural network features for machines with minimal rate penalty."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2512.08227/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":18,"sample":[{"doi":"","year":2025,"title":"New VVC profiles targeting Feature Coding for Machines","work_id":"76641795-3c40-4a46-8427-8670a233ec19","ref_index":1,"cited_arxiv_id":"2512.08227","is_internal_anchor":true},{"doi":"","year":null,"title":"FEATURE CODING FOR MACHINES Fig. 2 outlines the encoding and decoding process for in- termediate features computed by a neural network parti- tioned into NN Part-1 and NN Part-2.Xrepresents the set of","work_id":"0ca72932-ac3d-4742-8a03-8473bd36c10a","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"LOW-COMPLEXITY VVC PROFILES We evaluate VVC/H.266 for compressing intermediate fea- tures extracted from neural networks, with a focus on iden- tifying a lightweight yet effective tool-set for Feature","work_id":"3d476f8c-224d-4121-890f-978a3c2402ca","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"CONCLUSION We present a comprehensive analysis of VVC coding tools for compressing intermediate features in split-inference sys- tems. By profiling encoder decisions and conducting targeted ablation s","work_id":"ca8ed4a5-f4f7-4a40-89d4-049f8035202e","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"Deep feature com- pression for collaborative object detection","work_id":"c96c583b-574c-48c2-8049-e1b737c12939","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":18,"snapshot_sha256":"1e74b56f9bc270098e3aa454cb86c67986628b1e84597efe316566ba7484110d","internal_anchors":3},"formal_canon":{"evidence_count":2,"snapshot_sha256":"59c26c3b71fb474044915c27224ac2978042eebeeed4676309e2ff3cdfdd7f78"},"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":"2512.08227","created_at":"2026-07-15T00:21:15.234401+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.08227v1","created_at":"2026-07-15T00:21:15.234401+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.08227","created_at":"2026-07-15T00:21:15.234401+00:00"},{"alias_kind":"pith_short_12","alias_value":"YJCNZ7CTLVHN","created_at":"2026-07-15T00:21:15.234401+00:00"},{"alias_kind":"pith_short_16","alias_value":"YJCNZ7CTLVHNWRFV","created_at":"2026-07-15T00:21:15.234401+00:00"},{"alias_kind":"pith_short_8","alias_value":"YJCNZ7CT","created_at":"2026-07-15T00:21:15.234401+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2512.08227","citing_title":"New VVC profiles targeting Feature Coding for Machines","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U","json":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U.json","graph_json":"https://pith.science/api/pith-number/YJCNZ7CTLVHNWRFVSLYTGLEH5U/graph.json","events_json":"https://pith.science/api/pith-number/YJCNZ7CTLVHNWRFVSLYTGLEH5U/events.json","paper":"https://pith.science/paper/YJCNZ7CT"},"agent_actions":{"view_html":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U","download_json":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U.json","view_paper":"https://pith.science/paper/YJCNZ7CT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.08227&json=true","fetch_graph":"https://pith.science/api/pith-number/YJCNZ7CTLVHNWRFVSLYTGLEH5U/graph.json","fetch_events":"https://pith.science/api/pith-number/YJCNZ7CTLVHNWRFVSLYTGLEH5U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U/action/storage_attestation","attest_author":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U/action/author_attestation","sign_citation":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U/action/citation_signature","submit_replication":"https://pith.science/pith/YJCNZ7CTLVHNWRFVSLYTGLEH5U/action/replication_record"}},"created_at":"2026-07-15T00:21:15.234401+00:00","updated_at":"2026-07-15T00:21:15.234401+00:00"}