Bo Li
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
Bo Li has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Zhiheng Zhou
South China University of Technology
5 shared publications Han Liu
South China University of Technology
3 shared publicationsHang Gao
South China University of Technology
3 shared publicationsDelu Zeng
2 shared publications- Delu Zeng
School of Electronic and Information Engineering, South China University of Technology, Guangzhou, China
2 shared publications Tao Zheng
South China University of Technology
2 shared publications
Latest Publications
- Open Access
- 21 May 2026
- Published by Elsevier in SSRN Electronic Journal
See Fine, Know Clear: Boosting Semi-Supervised Video Object Segmentation via MambaVision-Powered Semantic and Fine-Grained Cues
Memory-bank-based methods have achieved promising performance in semi-supervised video object segmentation (SVOS). However, most existing methods store only single-scale coarse-grained features in the memory bank, leading to suboptimal performance on fine-grained segmentation. Furthermore, memory matching is conducted solely at the pixel level, which lacks high-level semantic modeling of objects as holistic entities, thereby causing ambiguity in distinguishing visually similar objects. In this work, we propose an SVOS framework based on multi-scale memory representation and semantic enhancement. The proposed framework stores multi-scale fine-grained features in the memory bank and fuses instance-level semantic information extracted from MambaVision, simultaneously boosting sensitivity to fine-grained object details and mitigating ambiguity in distinguishing similar objects. We accordingly name the proposed model See Fine, Know Clear (SFKC). This design introduces a key challenge: elevated computational costs induced by fine-grained feature matching. To address this issue, we design a Top-K-based filtering mechanism that streamlines the matching process and substantially reduces computational overhead. In addition, we design a dynamic memory management module to provide stable and effective reference features for subsequent frame segmentation. Extensive experiments demonstrate that our method achieves accurate and robust foreground object segmentation while retaining high efficiency when incorporating fine-grained memory features.
- 21 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSee Fine, Know Clear: Boosting Semi-Supervised Video Object Segmentation via MambaVision-Powered Semantic and Fine-Grained Cues
Memory-bank-based methods have achieved promising performance in semi-supervised video object segmentation (SVOS). However, most existing methods store only single-scale coarse-grained features in the memory bank, leading to suboptimal performance on fine-grained segmentation. Furthermore, memory matching is conducted solely at the pixel level, which lacks high-level semantic modeling of objects as holistic entities, thereby causing ambiguity in distinguishing visually similar objects. In this work, we propose an SVOS framework based on multi-scale memory representation and semantic enhancement. The proposed framework stores multi-scale fine-grained features in the memory bank and fuses instance-level semantic information extracted from MambaVision, simultaneously boosting sensitivity to fine-grained object details and mitigating ambiguity in distinguishing similar objects. We accordingly name the proposed model See Fine, Know Clear (SFKC). This design introduces a key challenge: elevated computational costs induced by fine-grained feature matching. To address this issue, we design a Top-K-based filtering mechanism that streamlines the matching process and substantially reduces computational overhead. In addition, we design a dynamic memory management module to provide stable and effective reference features for subsequent frame segmentation. Extensive experiments demonstrate that our method achieves accurate and robust foreground object segmentation while retaining high efficiency when incorporating fine-grained memory features.
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 3038-3053
Cross-View and Multi-Step Interaction for Change Captioning
Change captioning is a task that describes changes in image pairs using natural language. This task is more complex than single-image captioning as it requires a comprehensive understanding of each image and the ability to recognize and describe the semantic changes in image pairs. The key challenge lies in making the network generate an accurate and stable change representation under the interference of viewpoint shift. In this paper, we propose a cross-view and multi-step interaction network to generate robust change representation to resist pseudo-change. Specifically, in the intra-image representation learning stage, a cross-view interaction encoder is designed to enhance internal relationships by cross-referencing in image pairs. In the change feature learning stage, a multi-step change perceptron is employed to capture the change semantics from coarse to fine progressively. Then, a fusion module dynamically combines them as a fine-grained change representation. Besides, we propose a backward representation reconstruction module that facilitates the capture of semantic changes, thus improving the quality of captions in a self-supervised manner. Extensive experiments have shown that the method effectively captures real semantic changes under the interference of viewpoint shift and achieves state-of-the-art performance on five public datasets. The code is available at https://github.com/TTXiann/CVMSI
- 5 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 28 (15209210) , 3038-3053
Cross-View and Multi-Step Interaction for Change Captioning
Change captioning is a task that describes changes in image pairs using natural language. This task is more complex than single-image captioning as it requires a comprehensive understanding of each image and the ability to recognize and describe the semantic changes in image pairs. The key challenge lies in making the network generate an accurate and stable change representation under the interference of viewpoint shift. In this paper, we propose a cross-view and multi-step interaction network to generate robust change representation to resist pseudo-change. Specifically, in the intra-image representation learning stage, a cross-view interaction encoder is designed to enhance internal relationships by cross-referencing in image pairs. In the change feature learning stage, a multi-step change perceptron is employed to capture the change semantics from coarse to fine progressively. Then, a fusion module dynamically combines them as a fine-grained change representation. Besides, we propose a backward representation reconstruction module that facilitates the capture of semantic changes, thus improving the quality of captions in a self-supervised manner. Extensive experiments have shown that the method effectively captures real semantic changes under the interference of viewpoint shift and achieves state-of-the-art performance on five public datasets. The code is available at https://github.com/TTXiann/CVMSI
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-15
Dynamic and Asymmetric Enhancement for Remote Sensing Image Change Captioning
Remote Sensing Image Change Captioning (RSICC) plays a critical role in automated environmental monitoring by generating natural language descriptions that provide intuitive interpretations of changes between bi-temporal remote sensing images. Despite recent advancements, existing methods suffer from two fundamental limitations: (1) static encoding architectures fail to account for scene complexity and semantic diversity during feature extraction, leading to inflexible representation learning; and (2) symmetric computational structures are inherently unsuitable for modeling the asymmetric temporal dependencies inherent in “before-to-after” image pairs. To address these challenges, we propose a Dynamic Asymmetric Encoder (DAE), which introduces two key innovations. First, we design a difference-guided dynamic convolution module that adaptively adjusts convolutional parameters using input-driven scaling factors and offsets, thereby enabling scene-aware intra-image feature enhancement. Second, we develop a Multi-expert Temporal Interaction (METI) module that establishes an asymmetric computational topology: the “after” image branch actively perceives change information relative to the “before” image through three heterogeneous interaction experts, followed by feature fusion. This design allocates greater computational capacity to the “after” image branch while preserving temporal coherence. Furthermore, we introduce a Multi-level Feature Aggregator (MFA) that enhances the representation of salient changed regions across multiple scales via an iterative reinforcement mechanism. Experimental results validate the effectiveness of the proposed method, demonstrating state-of-the-art performance on two benchmark RSICC datasets. We publicly release our code repository at https://github.com/TTXiann/Dynamic-Asymmetric to facilitate future research.
- 21 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 63 (01962892) , 1-15
Dynamic and Asymmetric Enhancement for Remote Sensing Image Change Captioning
Remote Sensing Image Change Captioning (RSICC) plays a critical role in automated environmental monitoring by generating natural language descriptions that provide intuitive interpretations of changes between bi-temporal remote sensing images. Despite recent advancements, existing methods suffer from two fundamental limitations: (1) static encoding architectures fail to account for scene complexity and semantic diversity during feature extraction, leading to inflexible representation learning; and (2) symmetric computational structures are inherently unsuitable for modeling the asymmetric temporal dependencies inherent in “before-to-after” image pairs. To address these challenges, we propose a Dynamic Asymmetric Encoder (DAE), which introduces two key innovations. First, we design a difference-guided dynamic convolution module that adaptively adjusts convolutional parameters using input-driven scaling factors and offsets, thereby enabling scene-aware intra-image feature enhancement. Second, we develop a Multi-expert Temporal Interaction (METI) module that establishes an asymmetric computational topology: the “after” image branch actively perceives change information relative to the “before” image through three heterogeneous interaction experts, followed by feature fusion. This design allocates greater computational capacity to the “after” image branch while preserving temporal coherence. Furthermore, we introduce a Multi-level Feature Aggregator (MFA) that enhances the representation of salient changed regions across multiple scales via an iterative reinforcement mechanism. Experimental results validate the effectiveness of the proposed method, demonstrating state-of-the-art performance on two benchmark RSICC datasets. We publicly release our code repository at https://github.com/TTXiann/Dynamic-Asymmetric to facilitate future research.
- 1 August 2024
- Published by Springer Nature
- p. 3-14
Long-Short-Term Expert Attention Neural Networks for Traffic Flow Prediction
- 1 August 2024
- Published by Springer Nature
- p. 3-14
Long-Short-Term Expert Attention Neural Networks for Traffic Flow Prediction
- 22 September 2023
- Published by Springer Nature
- p. 501-513
Self-supervised Video Object Segmentation Using Motion Feature Compensation
- 22 September 2023
- Published by Springer Nature
- p. 501-513
Self-supervised Video Object Segmentation Using Motion Feature Compensation
- 14 June 2022
- Published by Elsevier in Signal Processing: Image Communication
- Vol. 107
TMVOS: Triplet Matching for Efficient Video Object Segmentation
- 14 June 2022
- Published by Elsevier in Signal Processing: Image Communication
- Vol. 107
TMVOS: Triplet Matching for Efficient Video Object Segmentation
- 5 December 2021
- Published by Springer Nature
- p. 353-365
STA3DCNN: Spatial-Temporal Attention 3D Convolutional Neural Network for Citywide Crowd Flow Prediction
- 5 December 2021
- Published by Springer Nature
- p. 353-365
STA3DCNN: Spatial-Temporal Attention 3D Convolutional Neural Network for Citywide Crowd Flow Prediction
- Open Access
Disentangled High Quality Salient Object Detection
Aiming at discovering and locating most distinctive objects from visual scenes, salient object detection (SOD) plays an essential role in various computer vision systems. Coming to the era of high resolution, SOD methods are facing new challenges. The major limitation of previous methods is that they try to identify the salient regions and estimate the accurate objects boundaries simultaneously with a single regression task at low-resolution. This practice ignores the inherent difference between the two difficult problems, resulting in poor detection quality. In this paper, we propose a novel deep learning framework for high-resolution SOD task, which disentangles the task into a low-resolution saliency classification network (LRSCN) and a high-resolution refinement network (HRRN). As a pixel-wise classification task, LRSCN is designed to capture sufficient semantics at low-resolution to identify the definite salient, background and uncertain image regions. HRRN is a regression task, which aims at accurately refining the saliency value of pixels in the uncertain region to preserve a clear object boundary at high-resolution with limited GPU memory. It is worth noting that by introducing uncertainty into the training process, our HRRN can well address the high-resolution refinement task without using any high-resolution training data. Extensive experiments on high-resolution saliency datasets as well as some widely used saliency benchmarks show that the proposed method achieves superior performance compared to the state-of-the-art methods.
Open AccessDisentangled High Quality Salient Object Detection
Aiming at discovering and locating most distinctive objects from visual scenes, salient object detection (SOD) plays an essential role in various computer vision systems. Coming to the era of high resolution, SOD methods are facing new challenges. The major limitation of previous methods is that they try to identify the salient regions and estimate the accurate objects boundaries simultaneously with a single regression task at low-resolution. This practice ignores the inherent difference between the two difficult problems, resulting in poor detection quality. In this paper, we propose a novel deep learning framework for high-resolution SOD task, which disentangles the task into a low-resolution saliency classification network (LRSCN) and a high-resolution refinement network (HRRN). As a pixel-wise classification task, LRSCN is designed to capture sufficient semantics at low-resolution to identify the definite salient, background and uncertain image regions. HRRN is a regression task, which aims at accurately refining the saliency value of pixels in the uncertain region to preserve a clear object boundary at high-resolution with limited GPU memory. It is worth noting that by introducing uncertainty into the training process, our HRRN can well address the high-resolution refinement task without using any high-resolution training data. Extensive experiments on high-resolution saliency datasets as well as some widely used saliency benchmarks show that the proposed method achieves superior performance compared to the state-of-the-art methods.
- Open Access
- 3 April 2020
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 34 (07) , 10591-10598
Frame-Guided Region-Aligned Representation for Video Person Re-Identification
Pedestrians in videos are usually in a moving state, resulting in serious spatial misalignment like scale variations and pose changes, which makes the video-based person re-identification problem more challenging. To address the above issue, in this paper, we propose a Frame-Guided Region-Aligned model (FGRA) for discriminative representation learning in two steps in an end-to-end manner. Firstly, based on a frame-guided feature learning strategy and a non-parametric alignment module, a novel alignment mechanism is proposed to extract well-aligned region features. Secondly, in order to form a sequence representation, an effective feature aggregation strategy that utilizes temporal alignment score and spatial attention is adopted to fuse region features in the temporal and spatial dimensions, respectively. Experiments are conducted on benchmark datasets to demonstrate the effectiveness of the proposed method to solve the misalignment problem and the superiority of the proposed method to the existing video-based person re-identification methods.
- 3 April 2020
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 34 (07) , 10591-10598
Open AccessFrame-Guided Region-Aligned Representation for Video Person Re-Identification
Pedestrians in videos are usually in a moving state, resulting in serious spatial misalignment like scale variations and pose changes, which makes the video-based person re-identification problem more challenging. To address the above issue, in this paper, we propose a Frame-Guided Region-Aligned model (FGRA) for discriminative representation learning in two steps in an end-to-end manner. Firstly, based on a frame-guided feature learning strategy and a non-parametric alignment module, a novel alignment mechanism is proposed to extract well-aligned region features. Secondly, in order to form a sequence representation, an effective feature aggregation strategy that utilizes temporal alignment score and spatial attention is adopted to fuse region features in the temporal and spatial dimensions, respectively. Experiments are conducted on benchmark datasets to demonstrate the effectiveness of the proposed method to solve the misalignment problem and the superiority of the proposed method to the existing video-based person re-identification methods.
- Open Access
- 1 November 2019
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Transactions on Information and Systems
- Vol. E102.D (11) , 2230-2237
Improving Slice-Based Model for Person Re-ID with Multi-Level Representation and Triplet-Center Loss
Person Re-Identification has received extensive study in the past few years and achieves impressive progress. Recent outstanding methods extract discriminative features by slicing feature maps of deep neural network into several stripes. Still there have improvement on feature fusion and metric learning strategy which can help promote slice-based methods. In this paper, we propose a novel framework that is end-to-end trainable, called Multi-level Slice-based Network (MSN), to capture features both in different levels and body parts. Our model consists of a dual-branch network architecture, one branch for global feature extraction and the other branch for local ones. Both branches process multi-level features using pyramid feature alike module. By concatenating the global and local features, distinctive features are exploited and properly compared. Also, our proposed method creatively introduces a triplet-center loss to elaborate combined loss function, which helps train the joint-learning network. By demonstrating the comprehensive experiments on the mainstream evaluation datasets including Market-1501, DukeMTMC, CUHK03, it indicates that our proposed model robustly achieves excellent performance and outperforms many of existing approaches. For example, on DukeMTMC dataset in single-query mode, we obtain a great result of Rank-1/mAP =85.9%(+1.0%)/74.2%(+4.7%).
- 1 November 2019
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in IEICE Transactions on Information and Systems
- Vol. E102.D (11) , 2230-2237
Open AccessImproving Slice-Based Model for Person Re-ID with Multi-Level Representation and Triplet-Center Loss
Person Re-Identification has received extensive study in the past few years and achieves impressive progress. Recent outstanding methods extract discriminative features by slicing feature maps of deep neural network into several stripes. Still there have improvement on feature fusion and metric learning strategy which can help promote slice-based methods. In this paper, we propose a novel framework that is end-to-end trainable, called Multi-level Slice-based Network (MSN), to capture features both in different levels and body parts. Our model consists of a dual-branch network architecture, one branch for global feature extraction and the other branch for local ones. Both branches process multi-level features using pyramid feature alike module. By concatenating the global and local features, distinctive features are exploited and properly compared. Also, our proposed method creatively introduces a triplet-center loss to elaborate combined loss function, which helps train the joint-learning network. By demonstrating the comprehensive experiments on the mainstream evaluation datasets including Market-1501, DukeMTMC, CUHK03, it indicates that our proposed model robustly achieves excellent performance and outperforms many of existing approaches. For example, on DukeMTMC dataset in single-query mode, we obtain a great result of Rank-1/mAP =85.9%(+1.0%)/74.2%(+4.7%).