{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:EG6SOGE6FHURFOPN254JCAOWSD","short_pith_number":"pith:EG6SOGE6","schema_version":"1.0","canonical_sha256":"21bd27189e29e912b9edd7789101d690c6ae594a0d0d636adb7af4a68905bc29","source":{"kind":"arxiv","id":"2104.04687","version":3},"attestation_state":"computed","paper":{"title":"Learning from 2D: Contrastive Pixel-to-Point Knowledge Transfer for 3D Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ching-Yu Tseng, Chin-Tang Chen, Hung-Ting Su, Hung-Yueh Chiang, Winston H. Hsu, Yueh-Cheng Liu, Yu-Kai Huang, Zhe-Yu Liu","submitted_at":"2021-04-10T05:40:42Z","abstract_excerpt":"Most 3D neural networks are trained from scratch owing to the lack of large-scale labeled 3D datasets. In this paper, we present a novel 3D pretraining method by leveraging 2D networks learned from rich 2D datasets. We propose the contrastive pixel-to-point knowledge transfer to effectively utilize the 2D information by mapping the pixel-level and point-level features into the same embedding space. Due to the heterogeneous nature between 2D and 3D networks, we introduce the back-projection function to align the features between 2D and 3D to make the transfer possible. Additionally, we devise a"},"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":"2104.04687","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2021-04-10T05:40:42Z","cross_cats_sorted":[],"title_canon_sha256":"b3bbc043e9b46f6fb9479985f0532c3bc3f19df4f788f6c548cda67694036901","abstract_canon_sha256":"ed16fd15456a2893081f51f044cb8707bdccf75feaeb6d8767549395c2913471"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:43:40.235972Z","signature_b64":"RWzufCetoUByIsyUb2+6mlomfM6KYJ9tYd/dnVjSQ1t8mt5rZodBHJ2AJNAlg+CKiNpnRugvLyWBBV4SyyxVAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"21bd27189e29e912b9edd7789101d690c6ae594a0d0d636adb7af4a68905bc29","last_reissued_at":"2026-07-05T03:43:40.235493Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:43:40.235493Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning from 2D: Contrastive Pixel-to-Point Knowledge Transfer for 3D Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ching-Yu Tseng, Chin-Tang Chen, Hung-Ting Su, Hung-Yueh Chiang, Winston H. Hsu, Yueh-Cheng Liu, Yu-Kai Huang, Zhe-Yu Liu","submitted_at":"2021-04-10T05:40:42Z","abstract_excerpt":"Most 3D neural networks are trained from scratch owing to the lack of large-scale labeled 3D datasets. In this paper, we present a novel 3D pretraining method by leveraging 2D networks learned from rich 2D datasets. We propose the contrastive pixel-to-point knowledge transfer to effectively utilize the 2D information by mapping the pixel-level and point-level features into the same embedding space. Due to the heterogeneous nature between 2D and 3D networks, we introduce the back-projection function to align the features between 2D and 3D to make the transfer possible. Additionally, we devise a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.04687","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/2104.04687/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":"2104.04687","created_at":"2026-07-05T03:43:40.235551+00:00"},{"alias_kind":"arxiv_version","alias_value":"2104.04687v3","created_at":"2026-07-05T03:43:40.235551+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.04687","created_at":"2026-07-05T03:43:40.235551+00:00"},{"alias_kind":"pith_short_12","alias_value":"EG6SOGE6FHUR","created_at":"2026-07-05T03:43:40.235551+00:00"},{"alias_kind":"pith_short_16","alias_value":"EG6SOGE6FHURFOPN","created_at":"2026-07-05T03:43:40.235551+00:00"},{"alias_kind":"pith_short_8","alias_value":"EG6SOGE6","created_at":"2026-07-05T03:43:40.235551+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20189","citing_title":"HilDA: Hierarchical Distillation with Diffusion for Advancing Self-Supervised LiDAR Pre-training","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00514","citing_title":"Cross4D-JEPA: Dense Cross-modal Correspondence Distillation for 4D Point Cloud Representation Learning","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03334","citing_title":"Bridging the Dimensionality Gap: A Taxonomy and Survey of 2D Vision Model Adaptation for 3D Analysis","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09899","citing_title":"Hyperbolic Distillation: Geometry-Guided Cross-Modal Transfer for Robust 3D Object Detection","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18831","citing_title":"Feasibility of Indoor Frame-Wise Lidar Semantic Segmentation via Distillation from Visual Foundation Model","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD","json":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD.json","graph_json":"https://pith.science/api/pith-number/EG6SOGE6FHURFOPN254JCAOWSD/graph.json","events_json":"https://pith.science/api/pith-number/EG6SOGE6FHURFOPN254JCAOWSD/events.json","paper":"https://pith.science/paper/EG6SOGE6"},"agent_actions":{"view_html":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD","download_json":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD.json","view_paper":"https://pith.science/paper/EG6SOGE6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2104.04687&json=true","fetch_graph":"https://pith.science/api/pith-number/EG6SOGE6FHURFOPN254JCAOWSD/graph.json","fetch_events":"https://pith.science/api/pith-number/EG6SOGE6FHURFOPN254JCAOWSD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD/action/storage_attestation","attest_author":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD/action/author_attestation","sign_citation":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD/action/citation_signature","submit_replication":"https://pith.science/pith/EG6SOGE6FHURFOPN254JCAOWSD/action/replication_record"}},"created_at":"2026-07-05T03:43:40.235551+00:00","updated_at":"2026-07-05T03:43:40.235551+00:00"}