Kun Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Kun Li has a total of 127 co-authors. Here are the top co-authors by number of shared publications:
- Jingyu Yang
Tianjin University
30 shared publications - Yu-Kun Lai
Cardiff University, United Kingdom
21 shared publications - Jinsong Zhang
School of Computer Science, Big Data and Software, Fuzhou University, Fuzhou 350108, China
10 shared publications - Yebin Liu
Tsinghua University, Beijing, China
9 shared publications - Huanjing Yue
Tianjin University
8 shared publications - Jianhua Guo
Chinese Academy of Sciences
7 shared publications
Latest Publications
- 29 August 2026
- Published by Springer Nature in International Journal of Computer Vision
- Vol. 134 (9) , 414
Feed-Forward Multi-view Multi-person Reconstruction with Contrastive Human-Aware 3D Representation
- 29 August 2026
- Published by Springer Nature in International Journal of Computer Vision
- Vol. 134 (9) , 414
Feed-Forward Multi-view Multi-person Reconstruction with Contrastive Human-Aware 3D Representation
- Open Access
- 1 August 2026
- Published by KEAI Publishing in Fundamental Research
TGRHuman: Text-guided realistic 3D human generation via diffusion renderer
- 1 August 2026
- Published by KEAI Publishing in Fundamental Research
Open AccessTGRHuman: Text-guided realistic 3D human generation via diffusion renderer
- 1 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 10 (3) , 2482-2497
Efficient HDR Reconstruction From Real-World Raw Images
The growing prevalence of high-resolution displays on edge devices has created a pressing need for efficient high dynamic range (HDR) imaging algorithms. However, most existing HDR methods either struggle to deliver satisfactory visual quality or incur high computational and memory costs, limiting their applicability to high-resolution inputs (typically exceeding 12 megapixels). Furthermore, current HDR dataset collection approaches are often labor-intensive and inefficient. In this work, we explore a novel and practical solution for HDR reconstruction directly from raw sensor data, aiming to enhance both performance and deployability on mobile platforms. Our key insights are threefold: (1) we propose RepUNet, a lightweight and efficient HDR network leveraging structural re-parameterization for fast and robust inference; (2) we design a new computational raw HDR data formation pipeline and construct a new raw HDR dataset, RealRaw-HDR; (3) we design a plug-and-play motion alignment loss to suppress ghosting artifacts under constrained bandwidth conditions effectively. Our model contains fewer than 830 K parameters and takes less than 3 ms to process an image of 4 K resolution using one RTX 3090 GPU. While being highly efficient, our model also achieves comparable performance to state-of-the-art HDR methods in terms of PSNR, SSIM, and a color difference metric.
- 1 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 10 (3) , 2482-2497
Efficient HDR Reconstruction From Real-World Raw Images
The growing prevalence of high-resolution displays on edge devices has created a pressing need for efficient high dynamic range (HDR) imaging algorithms. However, most existing HDR methods either struggle to deliver satisfactory visual quality or incur high computational and memory costs, limiting their applicability to high-resolution inputs (typically exceeding 12 megapixels). Furthermore, current HDR dataset collection approaches are often labor-intensive and inefficient. In this work, we explore a novel and practical solution for HDR reconstruction directly from raw sensor data, aiming to enhance both performance and deployability on mobile platforms. Our key insights are threefold: (1) we propose RepUNet, a lightweight and efficient HDR network leveraging structural re-parameterization for fast and robust inference; (2) we design a new computational raw HDR data formation pipeline and construct a new raw HDR dataset, RealRaw-HDR; (3) we design a plug-and-play motion alignment loss to suppress ghosting artifacts under constrained bandwidth conditions effectively. Our model contains fewer than 830 K parameters and takes less than 3 ms to process an image of 4 K resolution using one RTX 3090 GPU. While being highly efficient, our model also achieves comparable performance to state-of-the-art HDR methods in terms of PSNR, SSIM, and a color difference metric.
- 22 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-11
FOF-X: Towards Real-time Detailed Human Reconstruction from a Single Image
We introduce FOF-X for real-time reconstruction of detailed human geometry from a single image. Balancing real-time speed against high-quality results is a persistent challenge, mainly due to the high computational demands of existing 3D representations. To address this, we propose Fourier Occupancy Field (FOF), an efficient 3D representation by learning the Fourier series. The core of FOF is to factorize a 3D occupancy field into a 2D vector field, retaining topology and spatial relationships within the 3D domain while facilitating compatibility with 2D convolutional neural networks. Such a representation bridges the gap between 3D and 2D domains, enabling the integration of human parametric models as priors and enhancing the reconstruction robustness. Based on FOF, we design a new reconstruction framework, FOF-X, to avoid the performance degradation caused by texture and lighting. This enables our real-time reconstruction system to better handle the domain gap between training images and real images. Additionally, in FOF-X, we enhance the inter-conversion algorithms between FOF and mesh representations with a Laplacian constraint and an automaton-based discontinuity matcher, improving both quality and robustness. We validate the strengths of our approach on different datasets and real-captured data, where FOF-X achieves new state-of-the-art results.
- 22 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-11
FOF-X: Towards Real-time Detailed Human Reconstruction from a Single Image
We introduce FOF-X for real-time reconstruction of detailed human geometry from a single image. Balancing real-time speed against high-quality results is a persistent challenge, mainly due to the high computational demands of existing 3D representations. To address this, we propose Fourier Occupancy Field (FOF), an efficient 3D representation by learning the Fourier series. The core of FOF is to factorize a 3D occupancy field into a 2D vector field, retaining topology and spatial relationships within the 3D domain while facilitating compatibility with 2D convolutional neural networks. Such a representation bridges the gap between 3D and 2D domains, enabling the integration of human parametric models as priors and enhancing the reconstruction robustness. Based on FOF, we design a new reconstruction framework, FOF-X, to avoid the performance degradation caused by texture and lighting. This enables our real-time reconstruction system to better handle the domain gap between training images and real images. Additionally, in FOF-X, we enhance the inter-conversion algorithms between FOF and mesh representations with a Laplacian constraint and an automaton-based discontinuity matcher, improving both quality and robustness. We validate the strengths of our approach on different datasets and real-captured data, where FOF-X achieves new state-of-the-art results.
- Open Access
- 15 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-4
DreamCoser: Controllable Layered 3D Character Generation and Editing
- 15 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-4
Open AccessDreamCoser: Controllable Layered 3D Character Generation and Editing
- Open Access
DyCrowd: Towards Dynamic Crowd Reconstruction from a Large-scene Video
3D reconstruction of dynamic crowds in large scenes has become increasingly important for applications such as city surveillance and crowd analysis. However, current works attempt to reconstruct 3D crowds from a static image, causing a lack of temporal consistency and inability to alleviate the typical impact caused by occlusions. In this paper, we propose DyCrowd, the first framework for spatio-temporally consistent 3D reconstruction of hundreds of individuals' poses, positions and shapes from a large-scene video. We design a coarse-to-fine group-guided motion optimization strategy for occlusion-robust crowd reconstruction in large scenes. To address temporal instability and severe occlusions, we further incorporate a VAE (Variational Autoencoder)-based human motion prior along with a segment-level group-guided optimization. The core of our strategy leverages collective crowd behavior to address long-term dynamic occlusions. By jointly optimizing the motion sequences of individuals with similar motion segments and combining this with the proposed Asynchronous Motion Consistency (AMC) loss, we enable high-quality unoccluded motion segments to guide the motion recovery of occluded ones, ensuring robust and plausible motion recovery even in the presence of temporal desynchronization and rhythmic inconsistencies. Additionally, in order to fill the gap of no existing well-annotated large-scene video dataset, we contribute a virtual benchmark dataset, VirtualCrowd, for evaluating dynamic crowd reconstruction from large-scene videos. Experimental results demonstrate that the proposed method achieves state-of-the-art performance in the large-scene dynamic crowd reconstruction task. The code and dataset will be available for research purposes.
Open AccessDyCrowd: Towards Dynamic Crowd Reconstruction from a Large-scene Video
3D reconstruction of dynamic crowds in large scenes has become increasingly important for applications such as city surveillance and crowd analysis. However, current works attempt to reconstruct 3D crowds from a static image, causing a lack of temporal consistency and inability to alleviate the typical impact caused by occlusions. In this paper, we propose DyCrowd, the first framework for spatio-temporally consistent 3D reconstruction of hundreds of individuals' poses, positions and shapes from a large-scene video. We design a coarse-to-fine group-guided motion optimization strategy for occlusion-robust crowd reconstruction in large scenes. To address temporal instability and severe occlusions, we further incorporate a VAE (Variational Autoencoder)-based human motion prior along with a segment-level group-guided optimization. The core of our strategy leverages collective crowd behavior to address long-term dynamic occlusions. By jointly optimizing the motion sequences of individuals with similar motion segments and combining this with the proposed Asynchronous Motion Consistency (AMC) loss, we enable high-quality unoccluded motion segments to guide the motion recovery of occluded ones, ensuring robust and plausible motion recovery even in the presence of temporal desynchronization and rhythmic inconsistencies. Additionally, in order to fill the gap of no existing well-annotated large-scene video dataset, we contribute a virtual benchmark dataset, VirtualCrowd, for evaluating dynamic crowd reconstruction from large-scene videos. Experimental results demonstrate that the proposed method achieves state-of-the-art performance in the large-scene dynamic crowd reconstruction task. The code and dataset will be available for research purposes.
- 8 May 2025
- Published by Springer Nature in Frontiers of Computer Science
- Vol. 19 (11) , 1-2
EMP3D: an emergency medical procedures 3D dataset with pose and shape
- 8 May 2025
- Published by Springer Nature in Frontiers of Computer Science
- Vol. 19 (11) , 1-2
EMP3D: an emergency medical procedures 3D dataset with pose and shape
- Open Access
- 16 April 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Intelligent Systems and Technology
- Vol. 16 (2) , 1-23
JASRNet: Learning Joint Adaptive Sampling and Reconstruction for Depth Sensing
Recent attempts to exploit irregular sampling strategies for depth sensing have shown prominent merits over the uniform rectangular sampling in terms of depth reconstruction quality, particularly at low sampling rates. However, the separate treatment of depth sampling and reconstruction did not enjoy potential merits of joint optimization. In this article, we propose a joint adaptive depth sampling and reconstruction network, named JASRNet , for the RGB-D sensing configuration, to simultaneously optimize both the sampling and reconstruction of the depth information in an end-to-end manner. The sampling sub-network infers the locations to sample according to the significance distribution generated from the associated RGB image without any prior information of the underlying depth maps. The depth reconstruction sub-network learns and then fuses global and local depth features with attention guidance, which helps to obtain more accurate depth reconstruction results at boundaries. A hybrid loss function is further proposed to promote sharp discontinuities of the reconstructed depth maps. The qualitative and quantitative results show that our method achieves better depth sensing quality than several state-of-the-art methods for various indoor and outdoor scenes.
- 16 April 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Intelligent Systems and Technology
- Vol. 16 (2) , 1-23
Open AccessJASRNet: Learning Joint Adaptive Sampling and Reconstruction for Depth Sensing
Recent attempts to exploit irregular sampling strategies for depth sensing have shown prominent merits over the uniform rectangular sampling in terms of depth reconstruction quality, particularly at low sampling rates. However, the separate treatment of depth sampling and reconstruction did not enjoy potential merits of joint optimization. In this article, we propose a joint adaptive depth sampling and reconstruction network, named JASRNet , for the RGB-D sensing configuration, to simultaneously optimize both the sampling and reconstruction of the depth information in an end-to-end manner. The sampling sub-network infers the locations to sample according to the significance distribution generated from the associated RGB image without any prior information of the underlying depth maps. The depth reconstruction sub-network learns and then fuses global and local depth features with attention guidance, which helps to obtain more accurate depth reconstruction results at boundaries. A hybrid loss function is further proposed to promote sharp discontinuities of the reconstructed depth maps. The qualitative and quantitative results show that our method achieves better depth sensing quality than several state-of-the-art methods for various indoor and outdoor scenes.
- 13 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 31 (10) , 6737-6750
SpeechAct: Towards Generating Whole-Body Motion From Speech
Whole-body motion generation from speech audio is crucial for computer graphics and immersive VR/AR. Prior methods struggle to produce natural and diverse whole-body motions from speech. In this paper, we introduce a novel method, named SpeechAct, based on a hybrid point representation and contrastive motion learning to boost realism and diversity in motion generation. Our hybrid point representation leverages the advantages of keypoint representation and surface points of 3D body model, which is easy to learn and helps to achieve smooth and natural motion generation from speech audio. We design a VQ-VAE to learn a motion codebook using our hybrid presentation, and then regress the motion from the input audio using a translation model. To boost diversity in motion generation, we propose a contrastive motion learning method according to the intuitive idea that the generated motion should be different from the motions of other audios and other speakers. We collect negative samples from other audio inputs and other speakers using our translation model. With these negative samples, we pull the current motion away from them using a contrastive loss to produce more distinctive representations. In addition, we compose a face generator to generate deterministic face motion due to the strong connection between the face movements and the speech audio. Experimental results validate the superior performance of our model. The code is available at http://cic.tju.edu.cn/faculty/likun/projects/SpeechAct/index.html.
- 13 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 31 (10) , 6737-6750
SpeechAct: Towards Generating Whole-Body Motion From Speech
Whole-body motion generation from speech audio is crucial for computer graphics and immersive VR/AR. Prior methods struggle to produce natural and diverse whole-body motions from speech. In this paper, we introduce a novel method, named SpeechAct, based on a hybrid point representation and contrastive motion learning to boost realism and diversity in motion generation. Our hybrid point representation leverages the advantages of keypoint representation and surface points of 3D body model, which is easy to learn and helps to achieve smooth and natural motion generation from speech audio. We design a VQ-VAE to learn a motion codebook using our hybrid presentation, and then regress the motion from the input audio using a translation model. To boost diversity in motion generation, we propose a contrastive motion learning method according to the intuitive idea that the generated motion should be different from the motions of other audios and other speakers. We collect negative samples from other audio inputs and other speakers using our translation model. With these negative samples, we pull the current motion away from them using a contrastive loss to produce more distinctive representations. In addition, we compose a face generator to generate deterministic face motion due to the strong connection between the face movements and the speech audio. Experimental results validate the superior performance of our model. The code is available at http://cic.tju.edu.cn/faculty/likun/projects/SpeechAct/index.html.
- 15 December 2024
- Published by Springer Nature in The Visual Computer
- Vol. 41 (8) , 5781-5798
PISE-V: person image and video synthesis with decoupled GAN
- 15 December 2024
- Published by Springer Nature in The Visual Computer
- Vol. 41 (8) , 5781-5798
PISE-V: person image and video synthesis with decoupled GAN