{"paper":{"title":"HumanSplatHMR: Closing the Loop Between Human Mesh Recovery and Gaussian Splatting Avatar","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Joint optimization refines 3D human poses by routing rendering losses back through a Gaussian splatting avatar.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Katherine A. Skinner, Pou-Chun Kung, Ram Vasudevan, Seth Isaacson, Yeheng Zong, Yike Pan, Yizhou Chen","submitted_at":"2026-05-04T16:26:11Z","abstract_excerpt":"Accurately recovering human pose and appearance from video is an essential component of scene reconstruction, with applications to motion capture, motion prediction, virtual reality, and digital twinning. Despite significant interest in building realistic human avatars from video, this paper demonstrates that existing methods do not accurately recover the 3D geometry of humans. ViT-based approaches are not consistently reliable and can overfit to 2D views, while NeRF- and Gaussian Splatting-based avatars treat pose and appearance separately, limiting rendering generalization to new poses. To r"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments show consistent improvements over pose recovery baselines that omit image-level refinement and avatar baselines that decouple pose estimation from avatar reconstruction.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That backpropagating photometric, segmentation, and depth losses through the differentiable renderer will reliably refine global 3D poses without introducing instability or local minima that degrade the avatar quality.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"HumanSplatHMR closes the loop between human mesh recovery and Gaussian Splatting by using photometric, segmentation, and depth losses to refine poses during avatar optimization.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Joint optimization refines 3D human poses by routing rendering losses back through a Gaussian splatting avatar.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"7cf6af2c136e20f6e00e9b62ea80984e5529729f7ca5839d735a3551736de696"},"source":{"id":"2605.02784","kind":"arxiv","version":2},"verdict":{"id":"89b993de-7053-4e67-a7b3-3fdf7b4e83a1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T18:16:17.541804Z","strongest_claim":"Experiments show consistent improvements over pose recovery baselines that omit image-level refinement and avatar baselines that decouple pose estimation from avatar reconstruction.","one_line_summary":"HumanSplatHMR closes the loop between human mesh recovery and Gaussian Splatting by using photometric, segmentation, and depth losses to refine poses during avatar optimization.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That backpropagating photometric, segmentation, and depth losses through the differentiable renderer will reliably refine global 3D poses without introducing instability or local minima that degrade the avatar quality.","pith_extraction_headline":"Joint optimization refines 3D human poses by routing rendering losses back through a Gaussian splatting avatar."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.02784/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T15:33:59.307768Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T02:31:22.251119Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:59:12.561528Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"113355824d3d1456b452f519a3482aec78f009d75bf4339ca045a00a84803fb6"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":3,"snapshot_sha256":"25239003fadbd1e95e469711b7e4e8df7d5e8b49f77bbeeb10a577f01d7a7f3d"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}