pith. sign in

arxiv: 2606.06903 · v1 · pith:JK7N7OHRnew · submitted 2026-06-05 · 💻 cs.CV · cs.AI

Beyond Skeletons: Learning Animation Directly from Driving Videos with Same2X Training Strategy

Yuan Zeng , Yujia Shi , Yuhao Yang , Dongxia Liu , Zongqing Lu , Wenming Yang , Qingmin Liao This is my paper

Pith reviewed 2026-06-27 22:39 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords human animationvideo generationdriving videosdiffusion modelsidentity preservationocclusion robustnesstraining strategy
0
0 comments X

The pith

DirectAnimator generates animated videos by learning motion directly from driving videos rather than extracted poses.

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

The paper presents DirectAnimator as a way to animate a static reference image using information from a driving video without intermediate pose estimators that can fail under occlusion or complex movement. Instead, it extracts a Driving Cue Triplet of pose, face, and location cues and fuses them in a CueFusion DiT block to guide the generation process. A Same2X training strategy helps when the driving and reference persons differ by aligning their features to same-identity cases. If successful, this would make animation generation more reliable and less resource-intensive for creating videos of people in motion.

Core claim

DirectAnimator bypasses pose extraction and directly learns from raw driving videos. It introduces a Driving Cue Triplet that captures motion, expression, and alignment, fused via a CueFusion DiT block for control during denoising. The Same2X training strategy aligns cross-ID features with same-ID data to regularize optimization. Experiments show it achieves state-of-the-art visual quality and identity preservation, robust to occlusions and complex articulation, with fewer computational resources.

What carries the argument

The Driving Cue Triplet of pose, face, and location cues fused through the CueFusion DiT block, regularized by the Same2X training strategy that aligns cross-identity learning.

If this is right

  • Generated animations maintain higher visual quality and better identity preservation across different driving scenarios.
  • Performance remains stable even with occlusions or intricate body movements in the driving video.
  • Training and inference require fewer computational resources than methods relying on pose estimators.
  • Convergence during optimization is faster due to the regularization effect of Same2X training.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The direct-from-video approach could reduce errors propagated from imperfect pose detectors in other video synthesis tasks.
  • Such efficiency gains might allow deployment on consumer hardware for personalized animation creation.
  • Future extensions could adapt the cue triplet concept to non-human subjects or 3D animation.
  • By avoiding explicit skeletons, the method might handle stylistic or artistic driving videos more gracefully.

Load-bearing premise

The Driving Cue Triplet captures motion, expression, and alignment in a form stable enough to provide reliable control during denoising even across different identities.

What would settle it

Running the model on a set of driving videos with heavy occlusions or extreme poses and checking whether the output videos exhibit more artifacts or identity drift than a pose-based baseline.

Figures

Figures reproduced from arXiv: 2606.06903 by Dongxia Liu, Qingmin Liao, Wenming Yang, Yuan Zeng, Yuhao Yang, Yujia Shi, Zongqing Lu.

Figure 1
Figure 1. Figure 1: Our proposed driving cue provides a more robust representation for complex motions [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of DirectAnimator. (a) We replace the skeleton maps with our proposed driving [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Examples of driving cues. strategy, we discard low-quality segmentation results, forcing the model to rely on adjacent results for temporal reasoning. However, the segmented foreground contains rich appearance details (e.g., clothing and hair textures), such high-frequency information may distract the model from focusing on the pose information. Therefore, we apply low-pass filtering in the frequency domai… view at source ↗
Figure 4
Figure 4. Figure 4: (a) Comparison between different settings. In the Same-ID setting, the reference image and driving video share the same identity. In the more practical Cross-ID setting, they feature dif￾ferent identities. (b) Overview of the cross-ID training pipeline. First, a model is trained under the Same-ID setting. Then, in the Cross-ID training stage, a new model is trained using pseudo driving cues generated from … view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparisons between DirectAnimator and baselines on the TikTok (Row 1) [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sample results of DirectAnimator. User IDs are manually obscured for privacy protection. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: From left to right: original foreground color frame, grayscale frame with zoomed-in [PITH_FULL_IMAGE:figures/full_fig_p016_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Examples of pseudo driving cues. follows the driving pose. On the TikTok and Unseen test sets, we extract 2D body keypoints from both the driving and generated videos using DWpose, and compute the normalized distance between corresponding body landmarks. Facial Landmark Consistency (FLC) measures how accurately fa￾cial expressions and mouth shapes are transferred from the driving video. Similarly, we extra… view at source ↗
Figure 9
Figure 9. Figure 9: Qualitative comparisons with baseline methods, highlighting artifacts and showing the [PITH_FULL_IMAGE:figures/full_fig_p021_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Animation results of DirectAnimator, demonstrating (1) pose alignment, (2) identity [PITH_FULL_IMAGE:figures/full_fig_p022_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Failure cases of DirectAnimator. Case 1 shows loss of detail under mild motion blur, [PITH_FULL_IMAGE:figures/full_fig_p023_11.png] view at source ↗
read the original abstract

Human image animation aims to generate a video from a static reference image, guided by pose information extracted from a driving video. Existing approaches often rely on pose estimators to extract intermediate representations, but such signals are prone to errors under occlusion or complex poses. Building on these observations, we present DirectAnimator, a framework that bypasses pose extraction and directly learns from raw driving videos. We introduce a Driving Cue Triplet consisting of pose, face, and location cues that captures motion, expression, and alignment in a semantically rich yet stable form, and we fuse them through a CueFusion DiT block for reliable control during denoising. To make learning dependable when the driving and reference identities differ, we devise a Same2X training strategy that aligns cross-ID features with those learned from same-ID data, regularizing optimization and accelerating convergence. Extensive experiments demonstrate that DirectAnimator attains state-of-the-art visual quality and identity preservation while remaining robust to occlusions and complex articulation, and it does so with fewer computational resources. Our project page is at https://directanimator.github.io/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript proposes DirectAnimator, a framework for human image animation that bypasses pose estimators by learning animation signals directly from raw driving videos. It introduces a Driving Cue Triplet (pose, face, and location cues) fused through a CueFusion DiT block during denoising, along with a Same2X training strategy that aligns cross-identity features to same-identity data for regularization. The paper claims this yields state-of-the-art visual quality, identity preservation, robustness to occlusions and complex articulation, and reduced computational cost.

Significance. If the experimental claims hold, the work could meaningfully reduce dependence on error-prone intermediate pose representations in animation pipelines while providing a stable control mechanism via the cue triplet. The Same2X strategy addresses a practical cross-ID training challenge and may generalize to other conditional generation settings. No machine-checked proofs or parameter-free derivations are present.

major comments (2)
  1. [Abstract] Abstract: the central claims of state-of-the-art visual quality, identity preservation, robustness, and efficiency are asserted without any reported metrics, baselines, ablation studies, or quantitative results, leaving the primary empirical contribution unsupported by visible evidence.
  2. [Abstract] Abstract (and implied method): the Driving Cue Triplet and CueFusion DiT block are presented as enabling reliable control, yet no equations, extraction procedures, or architectural diagrams are supplied to allow verification that the cues are stable under occlusion or that the fusion mechanism avoids the very errors it claims to bypass.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We provide point-by-point responses to the major comments below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claims of state-of-the-art visual quality, identity preservation, robustness, and efficiency are asserted without any reported metrics, baselines, ablation studies, or quantitative results, leaving the primary empirical contribution unsupported by visible evidence.

    Authors: The abstract is intended as a concise summary of the work. Detailed quantitative results, including metrics, baselines, and ablations, are presented in the Experiments section (Section 5) of the manuscript. To better support the claims in the abstract, we will revise it to include key quantitative highlights from our experiments demonstrating the SOTA performance. revision: yes

  2. Referee: [Abstract] Abstract (and implied method): the Driving Cue Triplet and CueFusion DiT block are presented as enabling reliable control, yet no equations, extraction procedures, or architectural diagrams are supplied to allow verification that the cues are stable under occlusion or that the fusion mechanism avoids the very errors it claims to bypass.

    Authors: While the abstract offers a high-level overview, the full details are provided in the main text. Section 3.1 defines the Driving Cue Triplet and describes the extraction procedures for each cue. Section 3.2 presents the CueFusion DiT block with the corresponding equations for cue fusion during denoising. Figure 2 shows the architectural diagram. These sections explain how the approach maintains stability. No revision is required for this point. revision: no

Circularity Check

0 steps flagged

No significant circularity

full rationale

The abstract and available description introduce DirectAnimator, the Driving Cue Triplet, CueFusion DiT block, and Same2X strategy as novel components whose performance is asserted via experiments. No equations, self-citations, or derivations are supplied that reduce any claimed result to a fitted input or prior self-work by construction. The central claims remain framed as empirical outcomes rather than tautological redefinitions or renamings.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 3 invented entities

The central claim rests on the stability and sufficiency of the newly introduced cues and training strategy, which are postulated without external benchmarks or independent validation in the abstract.

axioms (1)
  • domain assumption Pose estimators are prone to errors under occlusion or complex poses
    Invoked in the opening observation to motivate bypassing pose extraction.
invented entities (3)
  • Driving Cue Triplet no independent evidence
    purpose: Captures motion, expression, and alignment from raw videos in a stable form
    New construct introduced to replace pose signals; no independent evidence provided.
  • CueFusion DiT block no independent evidence
    purpose: Fuses the three cues for reliable control during denoising
    New architectural component; no independent evidence provided.
  • Same2X training strategy no independent evidence
    purpose: Aligns cross-identity features with same-identity data to regularize optimization
    New training procedure; no independent evidence provided.

pith-pipeline@v0.9.1-grok · 5734 in / 1467 out tokens · 31450 ms · 2026-06-27T22:39:20.993311+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

33 extracted references · 11 linked inside Pith

  1. [1]

    Stable video diffusion: Scaling latent video diffusion models to large datasets.arXiv preprint arXiv:2311.15127,

    Andreas Blattmann, Tim Dockhorn, Sumith Kulal, Daniel Mendelevitch, Maciej Kilian, Dominik Lorenz, Yam Levi, Zion English, Vikram V oleti, Adam Letts, et al. Stable video diffusion: Scaling latent video diffusion models to large datasets.arXiv preprint arXiv:2311.15127,

    Pith/arXiv arXiv

  2. [2]

    X-dyna: Expressive dynamic human image animation

    Di Chang, Hongyi Xu, You Xie, Yipeng Gao, Zhengfei Kuang, Shengqu Cai, Chenxu Zhang, Guox- ian Song, Chao Wang, Yichun Shi, et al. X-dyna: Expressive dynamic human image animation. arXiv preprint arXiv:2501.10021,

    arXiv

  3. [3]

    Humandit: Pose-guided diffusion transformer for long-form human motion video generation.arXiv preprint arXiv:2502.04847,

    Qijun Gan, Yi Ren, Chen Zhang, Zhenhui Ye, Pan Xie, Xiang Yin, Zehuan Yuan, Bingyue Peng, and Jianke Zhu. Humandit: Pose-guided diffusion transformer for long-form human motion video generation.arXiv preprint arXiv:2502.04847,

    arXiv

  4. [4]

    Animatediff: Animate your personalized text-to-image diffu- sion models without specific tuning.arXiv preprint arXiv:2307.04725,

    Yuwei Guo, Ceyuan Yang, Anyi Rao, Zhengyang Liang, Yaohui Wang, Yu Qiao, Maneesh Agrawala, Dahua Lin, and Bo Dai. Animatediff: Animate your personalized text-to-image diffu- sion models without specific tuning.arXiv preprint arXiv:2307.04725,

    Pith/arXiv arXiv

  5. [5]

    No other representation component is needed: Diffusion transformers can provide representation guidance by themselves.arXiv preprint arXiv:2505.02831,

    Dengyang Jiang, Mengmeng Wang, Liuzhuozheng Li, Lei Zhang, Haoyu Wang, Wei Wei, Guang Dai, Yanning Zhang, and Jingdong Wang. No other representation component is needed: Diffusion transformers can provide representation guidance by themselves.arXiv preprint arXiv:2505.02831,

    arXiv

  6. [6]

    Repa-e: Unlocking vae for end-to-end tuning with latent diffusion transformers.arXiv preprint arXiv:2504.10483,

    Xingjian Leng, Jaskirat Singh, Yunzhong Hou, Zhenchang Xing, Saining Xie, and Liang Zheng. Repa-e: Unlocking vae for end-to-end tuning with latent diffusion transformers.arXiv preprint arXiv:2504.10483,

    arXiv

  7. [7]

    Anitalker: animate vivid and diverse talking faces through identity-decoupled facial motion encoding

    11 Published as a conference paper at ICLR 2026 Tao Liu, Feilong Chen, Shuai Fan, Chenpeng Du, Qi Chen, Xie Chen, and Kai Yu. Anitalker: animate vivid and diverse talking faces through identity-decoupled facial motion encoding. In Proceedings of the 32nd ACM International Conference on Multimedia, pp. 6696–6705,

    2026

  8. [8]

    Dreamactor-m1: Holistic, expressive and robust human image animation with hybrid guidance

    Yuxuan Luo, Zhengkun Rong, Lizhen Wang, Longhao Zhang, Tianshu Hu, and Yongming Zhu. Dreamactor-m1: Holistic, expressive and robust human image animation with hybrid guidance. arXiv preprint arXiv:2504.01724,

    arXiv

  9. [9]

    Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193,

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193,

    Pith/arXiv arXiv

  10. [10]

    Grounded sam: Assembling open-world models for diverse visual tasks.arXiv preprint arXiv:2401.14159,

    Tianhe Ren, Shilong Liu, Ailing Zeng, Jing Lin, Kunchang Li, He Cao, Jiayu Chen, Xinyu Huang, Yukang Chen, Feng Yan, et al. Grounded sam: Assembling open-world models for diverse visual tasks.arXiv preprint arXiv:2401.14159,

    Pith/arXiv arXiv

  11. [11]

    High- resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High- resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition, pp. 10684–10695, 2022a. Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High- resolutio...

    Pith/arXiv arXiv 2010

  12. [12]

    Stableanimator: High-quality identity-preserving human image animation.arXiv preprint arXiv:2411.17697,

    Shuyuan Tu, Zhen Xing, Xintong Han, Zhi-Qi Cheng, Qi Dai, Chong Luo, and Zuxuan Wu. Stableanimator: High-quality identity-preserving human image animation.arXiv preprint arXiv:2411.17697,

    arXiv

  13. [13]

    Towards accurate generative models of video: A new metric & challenges

    Thomas Unterthiner, Sjoerd Van Steenkiste, Karol Kurach, Raphael Marinier, Marcin Michalski, and Sylvain Gelly. Towards accurate generative models of video: A new metric & challenges. arXiv preprint arXiv:1812.01717,

    Pith/arXiv arXiv

  14. [14]

    Wan: Open and advanced large-scale video generative models.arXiv preprint arXiv:2503.20314,

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al. Wan: Open and advanced large-scale video generative models.arXiv preprint arXiv:2503.20314,

    Pith/arXiv arXiv

  15. [15]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024a

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024a. Tan Wang, Linjie Li, Kevin Lin, Yuanhao Zhai, Chung-Ching Lin, Zhengyuan Yang, Hanwang Zhang, Zicheng Liu,...

    Pith/arXiv arXiv 2026

  16. [16]

    Latent image animator: Learn- ing to animate images via latent space navigation.arXiv preprint arXiv:2203.09043,

    Yaohui Wang, Di Yang, Francois Bremond, and Antitza Dantcheva. Latent image animator: Learn- ing to animate images via latent space navigation.arXiv preprint arXiv:2203.09043,

    arXiv

  17. [17]

    X-portrait: Expres- sive portrait animation with hierarchical motion attention

    You Xie, Hongyi Xu, Guoxian Song, Chao Wang, Yichun Shi, and Linjie Luo. X-portrait: Expres- sive portrait animation with hierarchical motion attention. InACM SIGGRAPH 2024 Conference Papers, pp. 1–11,

    2024

  18. [18]

    Cogvideox: Text-to-video diffusion models with an expert transformer.arXiv preprint arXiv:2408.06072,

    Zhuoyi Yang, Jiayan Teng, Wendi Zheng, Ming Ding, Shiyu Huang, Jiazheng Xu, Yuanming Yang, Wenyi Hong, Xiaohan Zhang, Guanyu Feng, et al. Cogvideox: Text-to-video diffusion models with an expert transformer.arXiv preprint arXiv:2408.06072,

    Pith/arXiv arXiv

  19. [19]

    Representation alignment for generation: Training diffusion transformers is easier than you think.arXiv preprint arXiv:2410.06940,

    Sihyun Yu, Sangkyung Kwak, Huiwon Jang, Jongheon Jeong, Jonathan Huang, Jinwoo Shin, and Saining Xie. Representation alignment for generation: Training diffusion transformers is easier than you think.arXiv preprint arXiv:2410.06940,

    Pith/arXiv arXiv

  20. [20]

    Identity-preserving text-to-video generation by frequency decomposition.arXiv preprint arXiv:2411.17440,

    Shenghai Yuan, Jinfa Huang, Xianyi He, Yunyuan Ge, Yujun Shi, Liuhan Chen, Jiebo Luo, and Li Yuan. Identity-preserving text-to-video generation by frequency decomposition.arXiv preprint arXiv:2411.17440,

    arXiv

  21. [21]

    Mimicmotion: High-quality human motion video generation with confidence-aware pose guid- ance.arXiv preprint arXiv:2406.19680,

    Yuang Zhang, Jiaxi Gu, Li-Wen Wang, Han Wang, Junqi Cheng, Yuefeng Zhu, and Fangyuan Zou. Mimicmotion: High-quality human motion video generation with confidence-aware pose guid- ance.arXiv preprint arXiv:2406.19680,

    arXiv

  22. [22]

    Dynamictrl: Rethinking the basic structure and the role of text for high-quality human image animation.arXiv preprint arXiv:2503.21246,

    Haoyu Zhao, Zhongang Qi, Cong Wang, Qingping Zheng, Guansong Lu, Fei Chen, Hang Xu, and Zuxuan Wu. Dynamictrl: Rethinking the basic structure and the role of text for high-quality human image animation.arXiv preprint arXiv:2503.21246,

    arXiv

  23. [23]

    Open-sora: Democratizing efficient video production for all

    Zangwei Zheng, Xiangyu Peng, Tianji Yang, Chenhui Shen, Shenggui Li, Hongxin Liu, Yukun Zhou, Tianyi Li, and Yang You. Open-sora: Democratizing efficient video production for all. arXiv preprint arXiv:2412.20404,

    Pith/arXiv arXiv

  24. [24]

    Realisdance: Equip controllable character animation with real- istic hands.arXiv preprint arXiv:2409.06202,

    Jingkai Zhou, Benzhi Wang, Weihua Chen, Jingqi Bai, Dongyang Li, Aixi Zhang, Hao Xu, Mingyang Yang, and Fan Wang. Realisdance: Equip controllable character animation with real- istic hands.arXiv preprint arXiv:2409.06202,

    arXiv

  25. [25]

    Champ: Controllable and consistent human image animation with 3d parametric guidance

    Shenhao Zhu, Junming Leo Chen, Zuozhuo Dai, Yinghui Xu, Xun Cao, Yao Yao, Hao Zhu, and Siyu Zhu. Champ: Controllable and consistent human image animation with 3d parametric guidance. arXiv preprint arXiv:2403.14781,

    arXiv

  26. [26]

    13 Published as a conference paper at ICLR 2026 A APPENDIX In this appendix, we first present the foundational concepts and diffusion-based architectures in Section A.1. Section A.2 then provides an in-depth description of our Driving Cue representation, including the effect of low-pass filtering on pose cues, how spatial alignment is learned from pseudo ...

    2026

  27. [27]

    present a fully Transformer-based backbone for diffusion models, replacing the conventional convolutional U-Net architecture. Built upon the latent space framework of Stable Diffusion (Rombach et al., 2022b), DiT processes image representations encoded by a fixed V AE encoder into low-dimensional fea- ture maps. These latent tensors are segmented into non...

    2026

  28. [28]

    emerges as a scalable and high-performing diffusion Transformer architecture tailored for long-duration, text-conditioned video generation. Built upon the Diffusion Transformer (DiT) backbone (Peebles & Xie, 2023), CogVideoX integrates several critical innovations that address longstanding challenges in temporal coherence and cross-modal alignment. To eff...

    2023

  29. [29]

    To further suppress redundant information, we apply a low-pass filter in the frequency domain to eliminate high-frequency image details

    to segment out the foreground human subject. To further suppress redundant information, we apply a low-pass filter in the frequency domain to eliminate high-frequency image details. The resulting foreground image is used as the Pose Cue. While most prior methods adopt 68 facial landmarks as the driving signal for expression transfer, such sparse represent...

    2025

  30. [30]

    to the pose and face masks in the driving video, aligning their spatial layout and scale with that of the reference identity. To prevent potential information leakage during training, we further apply a grid-based softening operation on the pose mask, blurring the mask boundaries while retaining the coarse silhouette. These aligned pose and face masks tog...

    2026

  31. [31]

    In addition to the data used for same-ID training, we collect an extra set of 1,000 images featuring diverse identities as the pseudo reference set

    and MimicMotion (Zhang et al., 2024), two of the most competitive human image animation methods to date. In addition to the data used for same-ID training, we collect an extra set of 1,000 images featuring diverse identities as the pseudo reference set. For each driving video sampled from the same-ID training set, we randomly select 0 to 3 images from the...

    2024

  32. [32]

    and compute the average cosine similarity between corresponding pairs.Face Temporal Similarity (FTS)measures how temporally consistent the facial appearance remains within a generated video. We compute face embeddings for each frame using ArcFace and average the cosine similarity between embeddings of adjacent frames.Pose Landmark Consistency (PLC)measure...

    2026

  33. [33]

    Third, low video quality also degrades performance. For example, poor lighting conditions as in Case 3(1) or low spatial resolution as in Case 3(2) make it difficult to accurately infer the subject’s motion, resulting in noticeably lower animation quality. A.7 LIMITATIONS ANDFUTUREWORK While DirectAnimator demonstrates strong performance across various be...

    2026