Pith. sign in

REVIEW 4 major objections 4 minor 39 references

RGB-D Video Generation for Improving Human-to-Robot Object Handover Prediction

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that synthesizing full-body RGB-D handover video with realistic depth noise and gaze conditioning can substitute for large real-world handover datasets when training robot intention prediction, and supports this with…

desk verdict Useful dataset, plausible pipeline, but the transfer claim is a 7/10 vs 6/10 difference with a threshold fit to the test set—worth reviewing, needs stronger evidence. read the letter →

arxiv 2608.13028 v1 pith:PR4PV6ZA submitted 2026-08-13 cs.CV cs.RO

Tianyu Sun , Zhoujie Fu , Zihui Gao , Bang Zhang , Guosheng Lin This is my paper · ORCID
classification cs.CVcs.RO
keywords human-robothandoverRGB-Dvideogenerationintentionanticipationdiffusionsim-to-realtransferdepthnoisesimulationgazeconditioningsociallyawarerobotics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that a robot can learn to read human handover intent from generated, not just recorded, RGB-D video. It proposes a pipeline that synthesizes full-body handover clips from pose and facial-gaze cues, then corrupts the depth channel with realistic sensor-style void noise, and shows that training on these augmented clips transfers to a physical robot with far fewer false triggers. The reason to care: if synthetic data can bridge the sim-to-real gap this way, handover-capable robots no longer depend on expensive, large-scale human capture, and they can react to whole-body social signals instead of just hand position.

What carries the argument

PassGen combines two conditioning streams on a Stable Video Diffusion backbone: a pose-guidance branch that injects per-frame skeletons, and a Temporal Face Encoder (TFE) that turns facial-embedding sequences into temporally consistent face tokens $\tilde{F}_t$, which are injected along with pose guidance into the U-Net via cross-attention. For depth, a morphology-based erosion uses boundary-width mapping $w(h)$ and a recorded noise distribution $N_0$ to imprint realistic void patterns on an otherwise smooth DepthCrafter estimate. Downstream, the Intention Gating module fuses gaze confidence with approach velocity: $S_{\mathrm{intent}}(t)=\sigma(w_g f_{\mathrm{gaze}}(\tilde{F}_t)+w_v v_{\mathrm{obj}}(t))$, with a monotone max constraint that latches the trigger and a threshold $\tau=0.80$.

What would settle it

Replace the morphological depth-noise step with clean DepthCrafter output while keeping the same PassGen RGB stream and Intention Gating, then run the same negative trials on the physical robot; if the false-trigger rate stays near 2/30, the sensor-noise simulation is not the load-bearing component for sim-to-real transfer. If it rises, that step is carrying the transfer.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that full-body social context plus sensor-realistic depth noise, delivered through generated video, is what makes handover-intention prediction transfer from training data to a real robot. PassGen-generated clips (2,875 of the 5,000 Hand2Bot pairs) improve animation quality over prior animation models; adding them to real training raises mean intention accuracy from 87.5% to 90.0%, lowers mean FPR from 22.8% to 13.6%, and raises the unseen-object intention success rate from 6/10 to 7/10; and the full Intention-Gated system on a UR5e triggers 54/60 intended handovers while cutting false triggers from 25/30 to 2/30. The paper frames the synthetic contribution as variance reduction and false-positive suppression, not as broad manipulation superiority.

Load-bearing premise

The load-bearing premise is that the recorded L515 noise distribution and the boundary-width erosion $w(h)$ used to corrupt synthetic depth faithfully capture the artifacts that affect downstream prediction; the paper itself frames this as an empirical heuristic rather than a validated sensor model.

Editorial extensions

If this is right

  • Adding PassGen-Syn sequences to real training improves downstream intention accuracy from 87.5% to 90.0% and cuts mean FPR from 22.8% to 13.6% without changing the prediction architecture.
  • A physical UR5e with Intention Gating initiates 54/60 intended handovers and reduces false triggers from 25/30 to 2/30, evidence that gaze-plus-approach gating suppresses ambient motion.
  • Synthetic depth that carries realistic L515 void patterns prevents overfitting to clean geometry; replacing it with smooth DepthCrafter output is not sufficient.
  • Handover intention can be anticipated earlier from full-body social cues than from hand-centric views alone, which is the premise for socially aware shared-workspace behavior.
  • The combination of timestamped presentation labels and 6-DoF grasp annotations enables proactive control rather than reactive grasping.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If face-gaze tokens carry most of the intention signal, the same gating could transfer to other social-robotic tasks such as robot-initiated handover, where the robot must detect willingness rather than intention.
  • The depth-noise recipe is a cheap, sensor-specific transfer tool: record a sensor's void distribution, fit a boundary-width erosion, and reuse it for any generative depth pipeline; a natural test is to apply it to a second depth sensor and measure downstream false-trigger rate.
  • Because the monotone max trigger latches on the highest score, the system is deliberately biased toward early triggering; in longer autonomous sessions a single early false positive cannot be undone within a trial, so frame-level FPR may understate long-horizon risk.
  • The seen/unseen gap (ISR 6/10 vs 7/10 over only ten trials) is framed by the paper as variance reduction; a stronger test would run more objects and seeds before treating synthetic augmentation as a general manipulation improvement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The manuscript introduces Hand2Bot, a dataset of 5,000 RGB-D human-to-robot handover videos (2,125 real, 2,875 synthesized), and PassGen, a generative pipeline that combines Stable Video Diffusion with an Intention-Aware Temporal Face Encoder and a morphological depth-noise editing strategy. The authors also propose an Intention Gating module that fuses gaze features and object-approach velocity to trigger robot grasps. Experiments include animation-quality benchmarks, ablations on an evaluation subset of Hand2Bot-Real, and physical trials on a UR5e platform with 10 objects. The central claim is that training on PassGen-augmented data enables robust zero-shot transfer and earlier intention anticipation relative to hand-centric baselines.

Significance. If the transfer claim were established, the paper would make a useful contribution: it provides a full-body RGB-D handover dataset with realistic depth noise, a generative pipeline conditioned on explicit social cues, and an evaluation on a physical robot. The manuscript also contains strengths worth noting: a real-world data collection with informed consent, a clear gating formulation, and an unusually candid limitation statement in Section 6.6 conceding the lack of statistical significance. However, the central claim currently rests on a 6/10-to-7/10 unseen-object success difference and on an FPR reduction that appears to be achieved by selecting the activation threshold on the evaluation subset. As presented, the evidence does not substantiate 'robust zero-shot transfer' or the 'earlier intention anticipation' promised in the abstract.

major comments (4)
  1. [§6.6, Table 6; §6.7, Table 4] The central claim that PassGen-augmented training yields robust zero-shot transfer is not supported by the reported statistics. In Table 6, the unseen-object ISR increases from 6/10 to 7/10 (N=10), a difference well within binomial sampling noise, and Section 6.6 explicitly concedes the absence of formal statistical significance across random seeds. Moreover, the FPR improvement from 22.8% to 13.6% is compromised by the threshold-selection procedure: Section 6.7 reports an offline sweep over tau on the evaluation subset, and the selected tau=0.80 yields a frame-level FPR of 13.6%, identical to the 'Full Module' value in Tables 4 and 6; no held-out calibration set is described. The headline FPR is therefore a fitted value rather than an independent prediction. To support the transfer claim, the authors need either a held-out threshold calibration protocol or a significance test over multiple seeds/trials, and the reporting in Tables 4 and 6 should distinguish fitted from predicted quantities.
  2. [§5.3 and §6.5] The abstract claims 'earlier intention anticipation' relative to hand-centric baselines, but no temporal metric is reported anywhere in the experiments. Section 6.5 reports only ISR and FTR, and Section 6.7 reports frame-level FPR and interactive deadlocks, but there is no quantitative measurement of anticipation latency, time-to-trigger relative to presentation onset, or early-warning margin. Without such a metric, the 'earlier' component of the central claim is unsupported. The authors should add a timing analysis, e.g., the distribution of trigger times before the end of the presentation period, for the IG-enabled and baseline configurations.
  3. [§4.2] The depth-noise simulation is explicitly labeled as an 'empirical, heuristic morphology-based approximation' rather than a validated L515 sensor-noise model. This matters because Section 7 states that sensor-specific noise simulation is critical for cross-modal policy transfer, and the sim-to-real explanation in Section 4.2 depends on N0 and w(h) faithfully representing the boundary-level voids of the physical sensor. The paper should validate the heuristic quantitatively, e.g., by comparing the spatial and statistical distribution of zero-depth boundary pixels in generated depth against held-out real L515 frames, and should test whether downstream gains persist when the augmented data is evaluated with a different depth sensor or a different scene layout. As written, the depth-editing step is a plausible but unvalidated component of the transfer claim.
  4. [§6.5.1 and Table 5] The trial counts are internally inconsistent. The text says the authors selected 10 distinct objects (5 seen and 5 unseen) and ran positive trials per object type, which would imply 100 positive trials if 'per object' is literal. Table 5, however, lists six rows (five named seen objects plus one aggregate unseen row) with an overall 54/60, implying 60 positive trials. The table also reports only one aggregate 'Unseen-' row rather than per-object results for five unseen objects. Please clarify the number of objects, the number of trials per object, and report all per-object and per-category results so readers can reproduce the evaluation and assess variance across objects.
minor comments (4)
  1. [§6.6] The text says augmented training 'significantly improves downstream tracking accuracy, reduces FPR, and enhances robustness,' but two sentences later it concedes the result is without formal statistical significance. Please align the wording with the stated limitation, or provide the statistical test that supports 'significantly.'
  2. [Table 5] The row header 'Unseen-7 / 10' appears to be a formatting artifact; the values should be separated clearly, e.g., 'Unseen 7/10 0/5 3/5', to avoid ambiguity.
  3. [Figure 2] The caption says 'We mask all facial regions and replace them with the corresponding visualized depth maps,' but it is unclear what is displayed in the figure and why facial regions are masked in grasp-pose annotation; please clarify.
  4. [§6.2] The animation test set is described as '250 clips from Hand2Bot-Real,' but no details are given on how these clips were split from the 2,125 real sequences, whether the split is person-disjoint or scene-disjoint, or whether the PassGen training used all other real clips. Please provide the split procedure to assess possible information leakage.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline FPR is threshold-tuned on the same evaluation subset, making the reported FPR reduction a fitted value rather than an independent prediction.

  1. fitted input called prediction [Section 6.7; Section 6.4/Table 4; Section 6.6/Table 6]
    "Our selected operational configuration ofτ=0.80marks the ideal value, securing an optimal balance by maximizing intent tracking sensitivity for smooth haptic transitions while strictly dampening the offline continuous frame-level FPR to 13.6%."

    Section 6.7 selects the activation threshold τ=0.80 by an offline sweep that reports 13.6% frame-level FPR at that threshold. The same 13.6% value is then reported as the Full Module's FPR on the Hand2Bot-Real evaluation subset (Table 4) and as the Real+Syn Mean FPR (Table 6). The paper does not describe a held-out calibration set for the sweep; on its face, the threshold is tuned on the same evaluation data used to report the headline FPR. The claimed FPR reduction from synthetic augmentation (22.8% to 13.6%) is therefore at least partially forced by choosing the threshold that yields 13.6%, rather than being an independent prediction.

full rationale

The main circularity is the threshold-selection loop: the paper tunes τ=0.80 on an offline sweep and then presents the resulting 13.6% FPR as the Full Module's and Real+Syn's performance. This makes the offline FPR reduction a fitted quantity. The PassGen generation itself is otherwise evaluated on held-out test clips against external baselines, and the depth-noise simulation is explicitly labeled a heuristic approximation, so those parts are not circular. The physical robot FTR of 2/30 is a genuine held-out measurement, though it uses the same tuned threshold. The unseen-object ISR improvement (7/10 vs 6/10) is not itself constructed by the threshold, but the paper's own admission of no formal significance and N=10 limits that claim independently. Overall, the central zero-shot-transfer narrative retains some independent content, but the headline FPR 'prediction' reduces by construction to the threshold sweep.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claims rest on a small number of hand-set parameters (fusion weights), a threshold fitted to the evaluation data, and a heuristic depth-noise model whose parameters are not disclosed. The main domain assumptions concern the predictive power of face cues, the representativeness of the recorded noise, and the sufficiency of a scalar fusion for safe gating. No new physical entities are introduced; the TFE module is an algorithmic construct with no independent external evidence.

free parameters (4)
  • w_g = 1.0
    Gaze term weight in intention score (Eq. 2), set by hand; no optimization or sensitivity sweep is reported for this weight.
  • w_v = 2.5 [s/m]
    Object velocity weight in intention score (Eq. 2), set by hand; rationale given as balancing a typical reaching motion, but not fitted or swept.
  • tau = 0.80
    Activation threshold for intention trigger (Eq. 3 context); selected via offline sensitivity sweep on the evaluation subset (Section 6.7), so it is fitted to the data used to report headline FPR.
  • Boundary width vectors w(h) and noise distribution N0 = Not specified
    Parameters of the morphological depth noise model (Section 4.2); recorded from the authors' real captures and not reported numerically, making the depth simulation non-reproducible and potentially dataset-specific.
assumptions (5)
  • domain assumption In H2R handover, facial micro-expressions and gaze direction precede physical reaching motions and are reliable precursors of handover intention.
    Invoked in Section 4.1, item 2 to motivate the Temporal Face Encoder; the paper cites [9,10] but provides no direct measurement that these cues are necessary or sufficient.
  • domain assumption The recorded L515 noise distribution N0 and boundary width mapping w(h) are representative of the noise that affects downstream handover perception in deployment.
    Used in Section 4.2 for depth editing; the paper admits the model is heuristic and not validated against held-out sensor data.
  • ad hoc to paper ArcFace embeddings encode the gaze and micro-expression information needed for intention discrimination.
    TFE (Section 4.1) feeds ArcFace features into temporal attention; ArcFace is trained for identity recognition, so using it as a gaze/intent signal is an unvalidated design choice.
  • domain assumption A single scalar fusion of gaze confidence and object approach velocity, with monotone latching and threshold tau, is sufficient to safely gate grasping in real shared workspaces.
    The Intention Gating design (Sections 5.2-5.3) assumes this simple fusion generalizes; only one physical platform and one sensor configuration are tested.
  • domain assumption SVD fine-tuned with LoRA on Hand2Bot-Real generalizes to unseen objects and scenes.
    Underlies the zero-shot transfer claim; supported only by 10 unseen-object trials in Section 6.5.
invented entities (1)
  • Intention-Aware Temporal Face Encoder (TFE)
    purpose: Extracts temporally consistent face tokens from ArcFace embeddings to inject gaze/facial cues into the video diffusion model.
    A new network module; its ability to encode gaze or micro-expressions is not validated outside the paper's own ablations, and ArcFace is not a gaze estimator.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RGB-D Video Generation for Improving Human-to-Robot Object Handover Prediction." pith.science (2026). https://pith.science/paper/PR4PV6ZA

@misc{pith2026260813028,
  author       = {Pith},
  title        = {Pith review of: RGB-D Video Generation for Improving Human-to-Robot Object Handover Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PR4PV6ZA}},
  note         = {Machine review of arXiv:2608.13028}
}
read the original abstract

Human-to-robot (H2R) object handover is a fundamental capability for human-robot collaboration, yet progress is hindered by the scarcity of large-scale, human-centric datasets and the significant sim-to-real gap. To address these challenges, we introduce Hand2Bot, an RGB-D video dataset that provides rich contextual information such as body posture and facial expressions, specifically collected for handover scenarios with real-world noise patterns. We further propose PassGen, a generative pipeline that leverages stable video diffusion and an Intention-Aware Temporal Face Encoder to synthesize realistic handover sequences while ensuring hand-object consistency. To bridge the sim-to-real gap, we implement a morphology-based depth editing strategy that replicates realistic sensor noise found in physical depth maps. Experimental evaluations demonstrate that our framework achieves high intention identification accuracy and low false trigger rates in both ablation studies and real-world deployment on a physical robot platform. Our results confirm that training on PassGen allows for robust zero-shot transfer and earlier intention anticipation compared to traditional hand-centric baselines, effectively enabling socially aware robotic behavior in shared workspaces.

Figures

Figures reproduced from arXiv: 2608.13028 by the authors.

Figure 1
Figure 1. Existing bottlenecks in current H2R handover datasets. (a1) and (a2) compare a noisy real-world depth map against a clean synthetic depth map. (b1) and (b2) demonstrate the restricted, hand-centric field of view in mainstream benchmarks. To bridge these gaps, our primary motivation is to construct a comprehensive dataset that cap￾tures the full-body context alongside physical sensor artifacts. However, scaling such … view at source ↗
Figure 2
Figure 2. Grasp Pose Annotation. We give two examples of the GraspNet-based annotation on our Hand2Bot dataset. We mask all facial regions and replace them with the corresponding visualized depth maps. Specifically, let R ∈ SO(3) represent the orientation matrix of a candidate grasp pose, and vz = [0, 0, −1]T denote the canonical downward vertical axis aligned with gravity. We evaluate the angular deviation θ = arccos(gz · vz… view at source ↗
Figure 3
Figure 3. Pipeline of our PassGen. Our method is based on an SVD backbone, with two branches of networks. Given a pair of reference images Iref and video Vref , the appearance feature from Iref and the temporal feature from Vref are operated by corresponding modules separately. We also elaborate on the structures of TFE and Pose Guidance (PG) blocks in the bottom right corner. All real-world data collection and experiments we… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results on Hand2Bot-Real dataset. We compare our method with other SOTA human video generation algorithms on different reference images and video frames from our Hand2Bot-Real dataset. Our PassGen shows a promising and stable performance compared with other…
Figure 5
Figure 5. Figure 5: Qualitative comparison of generative RGB-D sequences. Our PassGen pipeline effec￾tively reconstructs human-object interaction geometry from sparse L515 inputs while maintaining realistic sensor noise patterns compared to over-smoothed baselines like DepthCrafter [14]. …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 32 canonical work pages

  1. [1]

    W. Yang, C. Paxton, A. Mousavian, Y .-W. Chao, M. Cakmak, and D. Fox. Reactive human-to- robot handovers of arbitrary objects. InICRA, 2021

    work page 2021

  2. [2]

    Ortenzi, A

    V . Ortenzi, A. Cosgun, T. Pardi, W. P. Chan, E. Croft, and D. Kuli ´c. Object handovers: a review for robotics.IEEE Trans. Robot., 37(6):1855–1873, 2021

    work page 2021

  3. [3]

    C. Meng, T. Zhang, and T. lun Lam. Fast and comfortable interactive robot-to-human object handover. InIROS, 2022

    work page 2022

  4. [4]

    Christen, L

    S. Christen, L. Feng, W. Yang, Y .-W. Chao, O. Hilliges, and J. Song. Synh2r: Synthesizing hand-object motions for learning human-to-robot handovers. InICRA, 2024

    work page 2024

  5. [5]

    Huang and K.-T

    Y .-Y . Huang and K.-T. Song. Human-to-robot handover control of an autonomous mobile robot based on hand-masked object pose estimation.IEEE Robotics and Automation Letters, 2024

    work page 2024

  6. [6]

    Y .-W. Chao, C. Paxton, Y . Xiang, W. Yang, B. Sundaralingam, T. Chen, et al. Handoversim: A simulation framework and benchmark for human-to-robot object handovers. InICRA, 2022

    work page 2022

  7. [7]

    Z. Wang, J. Chen, Z. Chen, P. Xie, R. Chen, and L. Yi. Genh2r: learning generalizable human- to-robot handover via scalable simulation demonstration and imitation. InCVPR, 2024

    work page 2024

  8. [8]

    Y . Wang, J. Ye, C. Xiao, Y . Zhong, H. Tao, H. Yu, Y . Liu, J. Yu, and Y . Ma. Dexh2r: A benchmark for dynamic dexterous grasping in human-to-robot handover. InICCV, 2025

    work page 2025

Show all 39 references
  1. [9]

    Huang, Y .-J

    Z. Huang, Y .-J. Mun, F. C. Pouria, and K. Driggs-Campbell. Hierarchical intention tracking with switching trees for real-time adaptation to dynamic human intentions, 2025. arXiv:2506.07004

    2025 arXiv

  2. [10]

    W. Wang, R. Li, Y . Chen, Y . Sun, and Y . Jia. Predicting human intentions in human–robot hand-over tasks through multimodal learning.IEEE Trans. Autom. Sci. Eng., 19(3):2339– 2353, 2021

    2021

  3. [11]

    L. Qiu, X. Gu, P. Li, Q. Zuo, W. Shen, J. Zhang, et al. Lhm: Large animatable human recon- struction model from a single image in seconds.arXiv preprint arXiv:2503.10625, 2025

    2025 arXiv

  4. [12]

    E. V . Mascaro, D. Sliwowski, and D. Lee. Hoi4abot: Human-object interaction anticipation for human intention reading collaborative robots.arXiv preprint arXiv:2309.16524, 2023

    2023 arXiv

  5. [13]

    L. Yang, B. Kang, Z. Huang, Z. Zhao, X. Xu, J. Feng, and H. Zhao. Depth anything v2. NeurIPS, 2024

    2024

  6. [14]

    W. Hu, X. Gao, X. Li, S. Zhao, X. Cun, et al. Depthcrafter: Generating consistent long depth sequences for open-world videos. InCVPR, 2025

    2025

  7. [15]

    X. Yi, Z. Wu, Q. Shen, Q. Xu, P. Zhou, J.-H. Lim, S. Yan, X. Wang, and H. Zhang. Mvgamba: Unify 3d content generation as state space sequence modeling. 2024

    2024

  8. [16]

    T. Sun, D. Hu, Y . Dai, and G. Wang. Diffusion-based depth inpainting for transparent and reflective objects.IEEE Transactions on Circuits and Systems for Video Technology, 2024

    2024

  9. [17]

    L. Hu. Animate anyone: Consistent and controllable image-to-video synthesis for character animation. InCVPR, 2024

    2024

  10. [18]

    T. Sun, Z. Fu, B. Zhang, and G. Lin. Mvanimate: Enhancing character animation with multi- view optimization, 2026. URLhttps://arxiv.org/abs/2602.08753

    2026

  11. [19]

    L. Hu, G. Wang, Z. Shen, X. Gao, D. Meng, L. Zhuo, P. . Zhang, and L. Bo. Animate any- one 2: High-fidelity character image animation with environment affordance.arXiv preprint arXiv:2502.06145, 2025. 12

    2025 arXiv

  12. [20]

    Z. Xu, Z. Huang, J. Cao, Y . Zhang, X. Cun, Q. Shuai, et al. Anchorcrafter: Animate cyber- anchors saling your products via human-object interacting video generation.arXiv preprint arXiv:2411.17383, 2024

    2024 arXiv

  13. [21]

    Y . Pang, R. Shao, J. Zhang, H. Tu, Y . Liu, et al. Manivideo: Generating hand-object manipu- lation video with dexterous and generalizable grasping. InCVPR, 2025

    2025

  14. [22]

    Laplaza, A

    J. Laplaza, A. Garrell, F. Moreno-Noguer, and A. Sanfeliu. Context and intention for 3d human motion prediction: experimentation and user study in handover tasks. InRO-MAN, 2022

    2022

  15. [23]

    Y .-W. Chao, W. Yang, Y . Xiang, P. Molchanov, A. Handa, et al. Dexycb: A benchmark for capturing hand grasping of objects. InCVPR, 2021

    2021

  16. [24]

    R. Ye, W. Xu, Z. S. Xue, T. Tang, Y . Wang, and C. Lu. H2o: A benchmark for visual human- human object handover analysis. InICCV, 2021

    2021

  17. [25]

    Wiederhold, A

    N. Wiederhold, A. Megyeri, D. Paris, S. Banerjee, and N. Banerjee. Hoh: Markerless multi- modal human-object-human handover dataset with large object count.NeurIPS, 2023

    2023

  18. [26]

    H.-S. Fang, C. Wang, M. Gou, and C. Lu. Graspnet-1billion: A large-scale benchmark for general object grasping. InCVPR, 2020

    2020

  19. [27]

    Blattmann, T

    A. Blattmann, T. Dockhorn, S. Kulal, D. Mendelevitch, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets.arXiv preprint arXiv:2311.15127, 2023

    2023 arXiv

  20. [28]

    Zhang, J

    Y . Zhang, J. Gu, L.-W. Wang, H. Wang, J. Cheng, Y . Zhu, and F. Zou. Mimicmotion: High- quality human motion video generation with confidence-aware pose guidance. InICML, 2025

    2025

  21. [29]

    Z. Yang, A. Zeng, C. Yuan, and Y . Li. Effective whole-body pose estimation with two-stages distillation. InICCV, 2023

    2023

  22. [30]

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. InCVPR, 2019

    2019

  23. [31]

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, et al. Lora: Low-rank adaptation of large language models. 2022

    2022

  24. [32]

    S. Tan, B. Gong, X. Wang, S. Zhang, D. Zheng, R. Zheng, K. Zheng, J. Chen, and M. Yang. Animate-x: Universal character image animation with enhanced motion representation. In ICLR, 2025

    2025

  25. [33]

    Chang, Y

    D. Chang, Y . Shi, Q. Gao, H. Xu, J. Fu, G. Song, et al. Magicpose: realistic human poses and facial expressions retargeting with identity-aware diffusion. InICML, 2024

    2024

  26. [34]

    Z. Xu, J. Zhang, J. H. Liew, H. Yan, J.-W. Liu, C. Zhang, et al. Magicanimate: Temporally consistent human image animation using diffusion model. InCVPR, 2024

    2024

  27. [35]

    Karras, A

    J. Karras, A. Holynski, T.-C. Wang, and I. Kemelmacher-Shlizerman. Dreampose: Fashion image-to-video synthesis via stable diffusion. InICCV, 2023

    2023

  28. [36]

    S. Tu, Z. Xing, X. Han, Z.-Q. Cheng, Q. Dai, C. Luo, and Z. Wu. Stableanimator: High-quality identity-preserving human image animation. InCVPR, 2025

    2025

  29. [37]

    Balaji, M

    Y . Balaji, M. R. Min, B. Bai, R. Chellappa, and H. P. Graf. Conditional gan with discriminative filter generation for text-to-video synthesis. InIJCAI, 2019

    2019

  30. [38]

    Unterthiner, S

    T. Unterthiner, S. Van Steenkiste, K. Kurach, R. Marinier, et al. Towards accurate generative models of video: A new metric & challenges.arXiv preprint arXiv:1812.01717, 2018

    2018 arXiv

  31. [39]

    Y . Song, C. B. Liu, W. Mao, and M. Z. Shou. Mitty: Diffusion-based human-to-robot video generation.arXiv preprint arXiv:2512.17253, 2025. 13

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.