{ "schema_version": 1, "reviewed_on": "2026-09-22", "scope": "MotionCheck v2 protocol: temporal alignment, procedure compliance, simulated motion tracking", "status": "Literature and design review; no reported reproduction or benchmark result", "primary_source_count": 7, "sources": [ { "id": "tcc_2019", "title": "Temporal Cycle-Consistency Learning", "authors": "Debidatta Dwibedi, Yusuf Aytar, Jonathan Tompson, Pierre Sermanet, Andrew Zisserman", "year": 2019, "venue": "CVPR", "paper_url": "https://openaccess.thecvf.com/content_CVPR_2019/html/Dwibedi_Temporal_Cycle-Consistency_Learning_CVPR_2019_paper.html", "code_url": "https://github.com/google-research/google-research/tree/master/tcc", "verified_claim": "Learns frame embeddings through temporal cycle consistency for correspondence and action-phase tasks.", "design_use": "Alignment is a useful observable, not proof that every required step occurred.", "release_note": "Official code identified in the v1 review; this review did not install or execute its learned model.", "implementation_relation": "Conceptual background; not reproduced." }, { "id": "lav_2021", "title": "Learning by Aligning Videos in Time", "authors": "Sanjay Haresh, Sateesh Kumar, Huseyin Coskun, Shahram N. Syed, Andrey Konin, M. Zeeshan Zia, Quoc-Huy Tran", "year": 2021, "venue": "CVPR", "paper_url": "https://openaccess.thecvf.com/content/CVPR2021/papers/Haresh_Learning_by_Aligning_Videos_in_Time_CVPR_2021_paper.pdf", "project_url": "https://retrocausal.ai/learning-by-aligning-videos-in-time/", "code_url": "https://github.com/trquhuytin/LAV-CVPR21", "verified_claim": "Combines Soft-DTW alignment with temporal regularization for self-supervised representation learning.", "design_use": "Keep temporal discrimination and coverage visible alongside alignment cost.", "release_note": "Official code identified; no trained dance-compliance model executed. Publication year is 2021.", "implementation_relation": "Motivation only; a hand-built ordered matcher is not LAV reproduction." }, { "id": "finediving_2022", "title": "FineDiving: A Fine-Grained Dataset for Procedure-Aware Action Quality Assessment", "authors": "Jinglin Xu, Yongming Rao, Xumin Yu, Guangyi Chen, Jie Zhou, Jiwen Lu", "year": 2022, "venue": "CVPR", "paper_url": "https://openaccess.thecvf.com/content/CVPR2022/html/Xu_FineDiving_A_Fine-Grained_Dataset_for_Procedure-Aware_Action_Quality_Assessment_CVPR_2022_paper.html", "code_url": "https://github.com/xujinglin/FineDiving", "verified_claim": "Uses fine-grained diving procedures and query/exemplar step correspondence for action quality assessment.", "design_use": "Report local stage evidence rather than only a whole-clip scalar.", "release_note": "Official PyTorch TSA code; dataset access requests a release agreement. README evaluates checkpoints produced by training. Not run here.", "limitation": "Sport-specific scoring and annotations do not establish general dance instruction compliance." }, { "id": "captaincook4d_2024", "title": "CaptainCook4D: A Dataset for Understanding Errors in Procedural Activities", "authors": "Rohith Peddi et al.", "year": 2024, "venue": "NeurIPS Datasets and Benchmarks", "paper_url": "https://arxiv.org/abs/2312.14556", "project_url": "https://captaincook4d.github.io/captain-cook/", "code_url": "https://github.com/CaptainCook4D/", "verified_claim": "Provides real procedural recordings with normal/error steps and evaluates error recognition and multistep localization.", "design_use": "Freeze independent step/error annotations, including order and outcome, before predictions.", "release_note": "Official project links code, data downloader, task graphs and annotations; not downloaded or evaluated here.", "limitation": "Cooking is a different domain; object-state errors are not observable from skeletons alone." }, { "id": "deepmimic_2018", "title": "DeepMimic: Example-Guided Deep Reinforcement Learning of Physics-Based Character Skills", "authors": "Xue Bin Peng, Pieter Abbeel, Sergey Levine, Michiel van de Panne", "year": 2018, "venue": "ACM TOG / SIGGRAPH", "doi": "10.1145/3197517.3201311", "project_url": "https://xbpeng.github.io/projects/DeepMimic/", "paper_url": "https://xbpeng.github.io/projects/DeepMimic/DeepMimic_2018.pdf", "code_url": "https://github.com/xbpeng/DeepMimic", "current_author_code_url": "https://github.com/xbpeng/MimicKit", "verified_claim": "Learns physics-based control policies with reference imitation and task objectives.", "design_use": "Evaluate actual state tracking and perturbation recovery, separately from visual similarity.", "release_note": "Legacy repository is marked deprecated. Author-linked MimicKit provides implementation and pretrained-model testing instructions; not run here.", "limitation": "Four-hinge fixed-base feedforward/PD tracking is not DeepMimic, RL training, or humanoid balance." }, { "id": "amp_2021", "title": "AMP: Adversarial Motion Priors for Stylized Physics-Based Character Control", "authors": "Xue Bin Peng, Ze Ma, Pieter Abbeel, Sergey Levine, Angjoo Kanazawa", "year": 2021, "venue": "ACM TOG / SIGGRAPH", "doi": "10.1145/3450626.3459670", "project_url": "https://xbpeng.github.io/projects/AMP/", "code_url": "https://github.com/xbpeng/MimicKit", "verified_claim": "Learns an adversarial motion prior from unstructured motion clips to provide style rewards for physical control.", "design_use": "Motion naturalness and compliance with a specific ordered reference are distinct objectives.", "release_note": "Author project links legacy implementation and MimicKit; no AMP model trained or evaluated here.", "limitation": "An explicit PD controller has no learned adversarial prior." }, { "id": "mujoco_actuation_docs", "title": "MuJoCo Documentation: actuation model and MJCF actuators", "authors": "MuJoCo maintainers / Google DeepMind", "year": null, "type": "Official versioned software documentation", "docs_url": "https://mujoco.readthedocs.io/en/stable/XMLreference.html#actuator-motor", "related_urls": [ "https://mujoco.readthedocs.io/en/stable/XMLreference.html#actuator-position", "https://mujoco.readthedocs.io/en/stable/computation/index.html#actuation-model" ], "verified_claim": "Motor, position-servo, control-limit and force-limit semantics differ; simulation must report actual actuator and state quantities.", "design_use": "Use matched torque motors and caps for feedforward versus encoder-PD ablation, with no state teleportation.", "release_note": "Stable documentation is mutable; record installed MuJoCo version and model XML/hash in the run manifest.", "limitation": "Successful simulator control is not evidence of hardware execution or sim-to-real transfer." } ] }