{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:3POX4UYOQLKMDAR77BW3MLU6GE","short_pith_number":"pith:3POX4UYO","schema_version":"1.0","canonical_sha256":"dbdd7e530e82d4c1823ff86db62e9e3102df481dc6d815a59a226ffcb06b2313","source":{"kind":"arxiv","id":"2109.03144","version":2},"attestation_state":"computed","paper":{"title":"PP-OCRv2: Bag of Tricks for Ultra Lightweight OCR System","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bin Lu, Cheng Cui, Chenxia Li, Dianhai Yu, Jun Zhou, Qiwen Liu, Ruoyu Guo, Weiwei Liu, Xiaoguang Hu, Yanjun Ma, Yehua Yang, Yuning Du","submitted_at":"2021-09-07T15:24:40Z","abstract_excerpt":"Optical Character Recognition (OCR) systems have been widely used in various of application scenarios. Designing an OCR system is still a challenging task. In previous work, we proposed a practical ultra lightweight OCR system (PP-OCR) to balance the accuracy against the efficiency. In order to improve the accuracy of PP-OCR and keep high efficiency, in this paper, we propose a more robust OCR system, i.e. PP-OCRv2. We introduce bag of tricks to train a better text detector and a better text recognizer, which include Collaborative Mutual Learning (CML), CopyPaste, Lightweight CPUNetwork (LCNet"},"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":"2109.03144","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2021-09-07T15:24:40Z","cross_cats_sorted":[],"title_canon_sha256":"e55e915eb9bb43a33c6198b542b4876763d6d89a237c2fb489bc8a574435d3a3","abstract_canon_sha256":"f42e613e254f0ade830018d2188842656a9fe0ab1e6a7ee64f515525b2422900"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:21:59.623279Z","signature_b64":"42T3k/18estTz9yAEnP/47LwWn6NCqYaG7GotBtYUD1TB/uPqDOIwppMQTHpgZn4rx+YKBykfTdRnaPnHb4uBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dbdd7e530e82d4c1823ff86db62e9e3102df481dc6d815a59a226ffcb06b2313","last_reissued_at":"2026-07-05T03:21:59.622778Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:21:59.622778Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PP-OCRv2: Bag of Tricks for Ultra Lightweight OCR System","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bin Lu, Cheng Cui, Chenxia Li, Dianhai Yu, Jun Zhou, Qiwen Liu, Ruoyu Guo, Weiwei Liu, Xiaoguang Hu, Yanjun Ma, Yehua Yang, Yuning Du","submitted_at":"2021-09-07T15:24:40Z","abstract_excerpt":"Optical Character Recognition (OCR) systems have been widely used in various of application scenarios. Designing an OCR system is still a challenging task. In previous work, we proposed a practical ultra lightweight OCR system (PP-OCR) to balance the accuracy against the efficiency. In order to improve the accuracy of PP-OCR and keep high efficiency, in this paper, we propose a more robust OCR system, i.e. PP-OCRv2. We introduce bag of tricks to train a better text detector and a better text recognizer, which include Collaborative Mutual Learning (CML), CopyPaste, Lightweight CPUNetwork (LCNet"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2109.03144","kind":"arxiv","version":2},"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/2109.03144/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":"2109.03144","created_at":"2026-07-05T03:21:59.622844+00:00"},{"alias_kind":"arxiv_version","alias_value":"2109.03144v2","created_at":"2026-07-05T03:21:59.622844+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2109.03144","created_at":"2026-07-05T03:21:59.622844+00:00"},{"alias_kind":"pith_short_12","alias_value":"3POX4UYOQLKM","created_at":"2026-07-05T03:21:59.622844+00:00"},{"alias_kind":"pith_short_16","alias_value":"3POX4UYOQLKMDAR7","created_at":"2026-07-05T03:21:59.622844+00:00"},{"alias_kind":"pith_short_8","alias_value":"3POX4UYO","created_at":"2026-07-05T03:21:59.622844+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.13108","citing_title":"PP-OCRv6: From 1.5M to 34.5M Parameters, Surpassing Billion-Scale VLMs on OCR Tasks","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29905","citing_title":"StrucTab: A Structured Optimization Framework for Table Parsing","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2409.01704","citing_title":"General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2507.05595","citing_title":"PaddleOCR 3.0 Technical Report","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19905","citing_title":"ViBR: Automated Bug Replay from Video-based Reports using Vision-Language Models","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE","json":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE.json","graph_json":"https://pith.science/api/pith-number/3POX4UYOQLKMDAR77BW3MLU6GE/graph.json","events_json":"https://pith.science/api/pith-number/3POX4UYOQLKMDAR77BW3MLU6GE/events.json","paper":"https://pith.science/paper/3POX4UYO"},"agent_actions":{"view_html":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE","download_json":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE.json","view_paper":"https://pith.science/paper/3POX4UYO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2109.03144&json=true","fetch_graph":"https://pith.science/api/pith-number/3POX4UYOQLKMDAR77BW3MLU6GE/graph.json","fetch_events":"https://pith.science/api/pith-number/3POX4UYOQLKMDAR77BW3MLU6GE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE/action/storage_attestation","attest_author":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE/action/author_attestation","sign_citation":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE/action/citation_signature","submit_replication":"https://pith.science/pith/3POX4UYOQLKMDAR77BW3MLU6GE/action/replication_record"}},"created_at":"2026-07-05T03:21:59.622844+00:00","updated_at":"2026-07-05T03:21:59.622844+00:00"}