{"paper":{"title":"ViPE: Video Pose Engine for 3D Geometric Perception","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"ViPE estimates camera poses and near-metric depth maps from any raw video without calibration.","cross_cats":["cs.GR","cs.RO","eess.IV"],"primary_cat":"cs.CV","authors_text":"Aleksandr Korovko, Chen-Hsuan Lin, Dmitry Slepichev, Hesam Rabeti, Huan Ling, Jiahui Huang, Jiawei Ren, Joydeep Biswas, Jun Gao, Kevin Xie, Laura Leal-Taixe, Qunjie Zhou, Sanja Fidler, Tianchang Shen, Xuanchi Ren","submitted_at":"2025-08-12T18:39:13Z","abstract_excerpt":"Accurate 3D geometric perception is an important prerequisite for a wide range of spatial AI systems. While state-of-the-art methods depend on large-scale training data, acquiring consistent and precise 3D annotations from in-the-wild videos remains a key challenge. In this work, we introduce ViPE, a handy and versatile video processing engine designed to bridge this gap. ViPE efficiently estimates camera intrinsics, camera motion, and dense, near-metric depth maps from unconstrained raw videos. It is robust to diverse scenarios, including dynamic selfie videos, cinematic shots, or dashcams, a"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ViPE outperforms existing uncalibrated pose estimation baselines by 18%/50% on TUM/KITTI sequences and annotates approximately 96M frames with accurate camera poses and dense depth maps.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the engine produces reliable near-metric depth and accurate poses on diverse in-the-wild videos without per-video calibration or ground-truth supervision.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ViPE estimates camera intrinsics, motion, and dense near-metric depth from uncalibrated videos, outperforming baselines on TUM and KITTI while releasing annotations for 96M frames across real and generated videos.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ViPE estimates camera poses and near-metric depth maps from any raw video without calibration.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"61c187d4b97a8818f47918556c555791d95e0d9494de57e3f09883d9dd7864f2"},"source":{"id":"2508.10934","kind":"arxiv","version":1},"verdict":{"id":"42b6ed31-6f02-424b-bf91-8113f0e6d7f6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T16:36:23.572987Z","strongest_claim":"ViPE outperforms existing uncalibrated pose estimation baselines by 18%/50% on TUM/KITTI sequences and annotates approximately 96M frames with accurate camera poses and dense depth maps.","one_line_summary":"ViPE estimates camera intrinsics, motion, and dense near-metric depth from uncalibrated videos, outperforming baselines on TUM and KITTI while releasing annotations for 96M frames across real and generated videos.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the engine produces reliable near-metric depth and accurate poses on diverse in-the-wild videos without per-video calibration or ground-truth supervision.","pith_extraction_headline":"ViPE estimates camera poses and near-metric depth maps from any raw video without calibration."},"references":{"count":91,"sample":[{"doi":"","year":2025,"title":"Cosmos World Foundation Model Platform for Physical AI","work_id":"a2dba24c-318d-476a-8b21-4289c265810c","ref_index":1,"cited_arxiv_id":"2501.03575","is_internal_anchor":true},{"doi":"","year":2025,"title":"Cosmos-transfer1: Conditional world generation with adaptive multimodal control","work_id":"f6758fe8-a1a6-4b00-9094-edba697f4c67","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"L4P: Low-level 4D vision perception unified","work_id":"bd61ab0e-c22e-42a4-98c2-c3d9d91f1b5c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"ARKitScenes: A Diverse Real-World Dataset For 3D Indoor Scene Understanding Using Mobile RGB-D Data","work_id":"0ce910be-ca1c-44c7-b7b1-c5353759d85e","ref_index":4,"cited_arxiv_id":"2111.08897","is_internal_anchor":true},{"doi":"","year":2024,"title":"Depth Pro: Sharp Monocular Metric Depth in Less Than a Second","work_id":"0b67883b-1901-45f1-9d58-1ef7a928df23","ref_index":5,"cited_arxiv_id":"2410.02073","is_internal_anchor":true}],"resolved_work":91,"snapshot_sha256":"74eaf682a12148ca8b703638049077152f95852d79728d5d327bfa59603314bc","internal_anchors":9},"formal_canon":{"evidence_count":1,"snapshot_sha256":"18b533f88823abe8d9070541caf2fe0b151148bead9d4628fbb804db8165c227"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}