Li Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Li Chen has a total of 59 co-authors. Here are the top co-authors by number of shared publications:
- Guangtao Zhai
Shanghai Jiao Tong University, Shanghai, China
11 shared publications Zhiyong Gao
Shanghai Artificial Intelligence Laboratory, Shanghai, China
11 shared publications- Xiaoyun Zhang
Shanghai Jiao Tong University
6 shared publications - Xiongkuo Min
Shanghai Jiao Tong University, Shanghai, China
6 shared publications - Huiyu Duan
Shanghai Jiao Tong University
5 shared publications - Zhiyong Gao
Institute of Image Communication and Network Engineering, Shanghai Jiao Tong University, Shanghai, China
5 shared publications
Latest Publications
- Open Access
A Multi-Stage Optimization Framework for Deploying Learned Image Compression on FPGAs
Deep learning-based image compression (LIC) has achieved state-of-the-art rate-distortion (RD) performance, yet deploying these models on resource-constrained FPGAs remains a major challenge. This work presents a complete, multi-stage optimization framework to bridge the gap between high-performance floating-point models and efficient, hardware-friendly integer-based implementations. First, we address the fundamental problem of quantization-induced performance degradation. We propose a Dynamic Range-Aware Quantization (DRAQ) method that uses statistically-calibrated activation clipping and a novel weight regularization scheme to counteract the effects of extreme data outliers and large dynamic ranges, successfully creating a high-fidelity 8-bit integer model. Second, building on this robust foundation, we introduce two hardware-aware optimization techniques tailored for FPGAs. A progressive mixed-precision search algorithm exploits FPGA flexibility to assign optimal, non-uniform bit-widths to each layer, minimizing complexity while preserving performance. Concurrently, a channel pruning method, adapted to work with the Generalized Divisive Normalization (GDN) layers common in LIC, removes model redundancy by eliminating inactive channels. Our comprehensive experiments show that the foundational DRAQ method reduces the BD-rate overhead of a GDN-based model from $30\%$ to $6.3\%$. The subsequent hardware-aware optimizations further reduce computational complexity by over $20\%$ with negligible impact on RD performance, yielding a final model that is both state-of-the-art in efficiency and superior in quality to existing FPGA-based LIC implementations.Open AccessA Multi-Stage Optimization Framework for Deploying Learned Image Compression on FPGAs
Deep learning-based image compression (LIC) has achieved state-of-the-art rate-distortion (RD) performance, yet deploying these models on resource-constrained FPGAs remains a major challenge. This work presents a complete, multi-stage optimization framework to bridge the gap between high-performance floating-point models and efficient, hardware-friendly integer-based implementations. First, we address the fundamental problem of quantization-induced performance degradation. We propose a Dynamic Range-Aware Quantization (DRAQ) method that uses statistically-calibrated activation clipping and a novel weight regularization scheme to counteract the effects of extreme data outliers and large dynamic ranges, successfully creating a high-fidelity 8-bit integer model. Second, building on this robust foundation, we introduce two hardware-aware optimization techniques tailored for FPGAs. A progressive mixed-precision search algorithm exploits FPGA flexibility to assign optimal, non-uniform bit-widths to each layer, minimizing complexity while preserving performance. Concurrently, a channel pruning method, adapted to work with the Generalized Divisive Normalization (GDN) layers common in LIC, removes model redundancy by eliminating inactive channels. Our comprehensive experiments show that the foundational DRAQ method reduces the BD-rate overhead of a GDN-based model from $30\%$ to $6.3\%$. The subsequent hardware-aware optimizations further reduce computational complexity by over $20\%$ with negligible impact on RD performance, yielding a final model that is both state-of-the-art in efficiency and superior in quality to existing FPGA-based LIC implementations. - Open Access
Exploring bidirectional bounds for minimax-training of Energy-based models
Energy-based models (EBMs) estimate unnormalized densities in an elegant framework, but they are generally difficult to train. Recent work has linked EBMs to generative adversarial networks, by noting that they can be trained through a minimax game using a variational lower bound. To avoid the instabilities caused by minimizing a lower bound, we propose to instead work with bidirectional bounds, meaning that we maximize a lower bound and minimize an upper bound when training the EBM. We investigate four different bounds on the log-likelihood derived from different perspectives. We derive lower bounds based on the singular values of the generator Jacobian and on mutual information. To upper bound the negative log-likelihood, we consider a gradient penalty-like bound, as well as one based on diffusion processes. In all cases, we provide algorithms for evaluating the bounds. We compare the different bounds to investigate, the pros and cons of the different approaches. Finally, we demonstrate that the use of bidirectional bounds stabilizes EBM training and yields high-quality density estimation and sample generation.
Open AccessExploring bidirectional bounds for minimax-training of Energy-based models
Energy-based models (EBMs) estimate unnormalized densities in an elegant framework, but they are generally difficult to train. Recent work has linked EBMs to generative adversarial networks, by noting that they can be trained through a minimax game using a variational lower bound. To avoid the instabilities caused by minimizing a lower bound, we propose to instead work with bidirectional bounds, meaning that we maximize a lower bound and minimize an upper bound when training the EBM. We investigate four different bounds on the log-likelihood derived from different perspectives. We derive lower bounds based on the singular values of the generator Jacobian and on mutual information. To upper bound the negative log-likelihood, we consider a gradient penalty-like bound, as well as one based on diffusion processes. In all cases, we provide algorithms for evaluating the bounds. We compare the different bounds to investigate, the pros and cons of the different approaches. Finally, we demonstrate that the use of bidirectional bounds stabilizes EBM training and yields high-quality density estimation and sample generation.
- 13 May 2025
- Published by Springer Nature in International Journal of Computer Vision
- p. 1-22
Exploring Bidirectional Bounds for Minimax-Training of Energy-Based Models
- 13 May 2025
- Published by Springer Nature in International Journal of Computer Vision
- p. 1-22
Exploring Bidirectional Bounds for Minimax-Training of Energy-Based Models
- 20 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 46  (8) , 5852-5872
A Coding Framework and Benchmark Towards Low-Bitrate Video Understanding
Video compression is indispensable to most video analysis systems. Despite saving the transportation bandwidth, it also deteriorates downstream video understanding tasks, especially at low-bitrate settings. To systematically investigate this problem, we first thoroughly review the previous methods, revealing that three principles, i.e. , task-decoupled, label-free, and data-emerged semantic prior, are critical to a machine-friendly coding framework but are not fully satisfied so far. In this paper, we propose a traditional-neural mixed coding framework that simultaneously fulfills all these principles, by taking advantage of both traditional codecs and neural networks (NNs). On one hand, the traditional codecs can efficiently encode the pixel signal of videos but may distort the semantic information. On the other hand, highly non-linear NNs are proficient in condensing video semantics into a compact representation. The framework is optimized by ensuring that a transportation-efficient semantic representation of the video is preserved w.r.t. the coding procedure, which is spontaneously learned from unlabeled data in a self-supervised manner. The videos collaboratively decoded from two streams (codec and NN) are of rich semantics, as well as visually photo-realistic, empirically boosting several mainstream downstream video analysis task performances without any post-adaptation procedure. Furthermore, by introducing the attention mechanism and adaptive modeling scheme, the video semantic modeling ability of our approach is further enhanced. Fianlly, we build a low-bitrate video understanding benchmark with three downstream tasks on eight datasets, demonstrating the notable superiority of our approach. All codes, data, and models will be open-sourced for facilitating future research.
- 20 February 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 46  (8) , 5852-5872
A Coding Framework and Benchmark Towards Low-Bitrate Video Understanding
Video compression is indispensable to most video analysis systems. Despite saving the transportation bandwidth, it also deteriorates downstream video understanding tasks, especially at low-bitrate settings. To systematically investigate this problem, we first thoroughly review the previous methods, revealing that three principles, i.e. , task-decoupled, label-free, and data-emerged semantic prior, are critical to a machine-friendly coding framework but are not fully satisfied so far. In this paper, we propose a traditional-neural mixed coding framework that simultaneously fulfills all these principles, by taking advantage of both traditional codecs and neural networks (NNs). On one hand, the traditional codecs can efficiently encode the pixel signal of videos but may distort the semantic information. On the other hand, highly non-linear NNs are proficient in condensing video semantics into a compact representation. The framework is optimized by ensuring that a transportation-efficient semantic representation of the video is preserved w.r.t. the coding procedure, which is spontaneously learned from unlabeled data in a self-supervised manner. The videos collaboratively decoded from two streams (codec and NN) are of rich semantics, as well as visually photo-realistic, empirically boosting several mainstream downstream video analysis task performances without any post-adaptation procedure. Furthermore, by introducing the attention mechanism and adaptive modeling scheme, the video semantic modeling ability of our approach is further enhanced. Fianlly, we build a low-bitrate video understanding benchmark with three downstream tasks on eight datasets, demonstrating the notable superiority of our approach. All codes, data, and models will be open-sourced for facilitating future research.
- 3 February 2024
- Published by Springer Nature
- p. 512-525
Perceptual Quality Assessment of Omnidirectional Audio-Visual Signals
- 3 February 2024
- Published by Springer Nature
- p. 512-525
Perceptual Quality Assessment of Omnidirectional Audio-Visual Signals
- 10 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- Vol. 70  (2) , 620-636
EffiHDR: An Efficient Framework for HDRTV Reconstruction and Enhancement in UHD Systems
Recent advancements in SDRTV-to-HDRTV conversion have yielded impressive results in reconstructing high dynamic range television (HDRTV) videos from standard dynamic range television (SDRTV) videos. However, the practical applications of these techniques are limited for ultra-high definition (UHD) video systems due to their high computational and memory costs. In this paper, we propose EffiHDR, an efficient framework primarily operating in the downsampled space, effectively reducing the computational and memory demands. Our framework comprises a real-time SDRTV-to-HDRTV Reconstruction model and a plug-and-play HDRTV Enhancement model. The SDRTV-to-HDRTV Reconstruction model learns affine transformation coefficients instead of directly predicting output pixels to preserve high-frequency information and mitigate information loss caused by downsampling. It decomposes SDRTV-to-HDR mapping into pixel intensity-dependent and local-dependent affine transformations. The pixel intensity-dependent transformation leverages global contexts and pixel intensity conditions to transform SDRTV pixels to the HDRTV domain. The local-dependent transformation predicts affine coefficients based on local contexts, further enhancing dynamic range, local contrast, and color tone. Additionally, we introduce a plug-and-play HDRTV Enhancement model based on an efficient Transformer-based U-net, which enhances luminance and color details in challenging recovery scenarios. Experimental results demonstrate that our SDRTV-to-HDRTV Reconstruction model achieves real-time 4K conversion with impressive performance. When combined with the HDRTV Enhancement model, our approach outperforms state-of-the-art methods in performance and efficiency.
- 10 January 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Broadcasting
- Vol. 70  (2) , 620-636
EffiHDR: An Efficient Framework for HDRTV Reconstruction and Enhancement in UHD Systems
Recent advancements in SDRTV-to-HDRTV conversion have yielded impressive results in reconstructing high dynamic range television (HDRTV) videos from standard dynamic range television (SDRTV) videos. However, the practical applications of these techniques are limited for ultra-high definition (UHD) video systems due to their high computational and memory costs. In this paper, we propose EffiHDR, an efficient framework primarily operating in the downsampled space, effectively reducing the computational and memory demands. Our framework comprises a real-time SDRTV-to-HDRTV Reconstruction model and a plug-and-play HDRTV Enhancement model. The SDRTV-to-HDRTV Reconstruction model learns affine transformation coefficients instead of directly predicting output pixels to preserve high-frequency information and mitigate information loss caused by downsampling. It decomposes SDRTV-to-HDR mapping into pixel intensity-dependent and local-dependent affine transformations. The pixel intensity-dependent transformation leverages global contexts and pixel intensity conditions to transform SDRTV pixels to the HDRTV domain. The local-dependent transformation predicts affine coefficients based on local contexts, further enhancing dynamic range, local contrast, and color tone. Additionally, we introduce a plug-and-play HDRTV Enhancement model based on an efficient Transformer-based U-net, which enhances luminance and color details in challenging recovery scenarios. Experimental results demonstrate that our SDRTV-to-HDRTV Reconstruction model achieves real-time 4K conversion with impressive performance. When combined with the HDRTV Enhancement model, our approach outperforms state-of-the-art methods in performance and efficiency.
- Open Access
Audio-visual Saliency for Omnidirectional Videos
Visual saliency prediction for omnidirectional videos (ODVs) has shown great significance and necessity for omnidirectional videos to help ODV coding, ODV transmission, ODV rendering, etc.. However, most studies only consider visual information for ODV saliency prediction while audio is rarely considered despite its significant influence on the viewing behavior of ODV. This is mainly due to the lack of large-scale audio-visual ODV datasets and corresponding analysis. Thus, in this paper, we first establish the largest audio-visual saliency dataset for omnidirectional videos (AVS-ODV), which comprises the omnidirectional videos, audios, and corresponding captured eye-tracking data for three video sound modalities including mute, mono, and ambisonics. Then we analyze the visual attention behavior of the observers under various omnidirectional audio modalities and visual scenes based on the AVS-ODV dataset. Furthermore, we compare the performance of several state-of-the-art saliency prediction models on the AVS-ODV dataset and construct a new benchmark. Our AVS-ODV datasets and the benchmark will be released to facilitate future research.
Open AccessAudio-visual Saliency for Omnidirectional Videos
Visual saliency prediction for omnidirectional videos (ODVs) has shown great significance and necessity for omnidirectional videos to help ODV coding, ODV transmission, ODV rendering, etc.. However, most studies only consider visual information for ODV saliency prediction while audio is rarely considered despite its significant influence on the viewing behavior of ODV. This is mainly due to the lack of large-scale audio-visual ODV datasets and corresponding analysis. Thus, in this paper, we first establish the largest audio-visual saliency dataset for omnidirectional videos (AVS-ODV), which comprises the omnidirectional videos, audios, and corresponding captured eye-tracking data for three video sound modalities including mute, mono, and ambisonics. Then we analyze the visual attention behavior of the observers under various omnidirectional audio modalities and visual scenes based on the AVS-ODV dataset. Furthermore, we compare the performance of several state-of-the-art saliency prediction models on the AVS-ODV dataset and construct a new benchmark. Our AVS-ODV datasets and the benchmark will be released to facilitate future research.
- 29 October 2023
- Published by Springer Nature
- p. 365-378
Audio-Visual Saliency for Omnidirectional Videos
- 29 October 2023
- Published by Springer Nature
- p. 365-378
Audio-Visual Saliency for Omnidirectional Videos
- Open Access
Perceptual Quality Assessment of Omnidirectional Audio-visual Signals
Omnidirectional videos (ODVs) play an increasingly important role in the application fields of medical, education, advertising, tourism, etc. Assessing the quality of ODVs is significant for service-providers to improve the user's Quality of Experience (QoE). However, most existing quality assessment studies for ODVs only focus on the visual distortions of videos, while ignoring that the overall QoE also depends on the accompanying audio signals. In this paper, we first establish a large-scale audio-visual quality assessment dataset for omnidirectional videos, which includes 375 distorted omnidirectional audio-visual (A/V) sequences generated from 15 high-quality pristine omnidirectional A/V contents, and the corresponding perceptual audio-visual quality scores. Then, we design three baseline methods for full-reference omnidirectional audio-visual quality assessment (OAVQA), which combine existing state-of-the-art single-mode audio and video QA models via multimodal fusion strategies. We validate the effectiveness of the A/V multimodal fusion method for OAVQA on our dataset, which provides a new benchmark for omnidirectional QoE evaluation. Our dataset is available at https://github.com/iamazxl/OAVQA.
Open AccessPerceptual Quality Assessment of Omnidirectional Audio-visual Signals
Omnidirectional videos (ODVs) play an increasingly important role in the application fields of medical, education, advertising, tourism, etc. Assessing the quality of ODVs is significant for service-providers to improve the user's Quality of Experience (QoE). However, most existing quality assessment studies for ODVs only focus on the visual distortions of videos, while ignoring that the overall QoE also depends on the accompanying audio signals. In this paper, we first establish a large-scale audio-visual quality assessment dataset for omnidirectional videos, which includes 375 distorted omnidirectional audio-visual (A/V) sequences generated from 15 high-quality pristine omnidirectional A/V contents, and the corresponding perceptual audio-visual quality scores. Then, we design three baseline methods for full-reference omnidirectional audio-visual quality assessment (OAVQA), which combine existing state-of-the-art single-mode audio and video QA models via multimodal fusion strategies. We validate the effectiveness of the A/V multimodal fusion method for OAVQA on our dataset, which provides a new benchmark for omnidirectional QoE evaluation. Our dataset is available at https://github.com/iamazxl/OAVQA.
- 10 March 2023
- Published by Springer Nature
- p. 191-202
FPGA-Based Hardware Implementation of JPEG XS Encoder
- 10 March 2023
- Published by Springer Nature
- p. 191-202
FPGA-Based Hardware Implementation of JPEG XS Encoder