Junbo Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Junbo Wang has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Yuke Li
School of Software, Northwestern Polytechnical University, Xian, China
6 shared publications - Hongsong Wang
School of Computer Science and Engineering, Southeast University, Nanjing, China
5 shared publications - Yining Zhu
School of Computer Science, Northwestern Polytechnical University, Xian, China
4 shared publications - Liangyu Fu
School of Software, Northwestern Polytechnical University, Xi'an, China
3 shared publications - Qiangguo Jin
Northwestern Polytechnical University
3 shared publications - Xuecheng Wu
Xi'an Jiaotong University, Xi'an, China
3 shared publications
Latest Publications
- Open Access
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1674-1683
TDVR: Joint Text Disambiguation and Viewpoint Reasoning for Zero-Shot 3D Visual Grounding
- 9 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 1674-1683
Open AccessTDVR: Joint Text Disambiguation and Viewpoint Reasoning for Zero-Shot 3D Visual Grounding
- 30 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-11
RGS: A Unified Representation-Generation-Selection Framework for Knowledge-Based Visual Question Answering
Knowledge-based visual question answering (VQA) is a task that answers questions with additional knowledge beyond the image itself. Existing methods have either retrieved external knowledge bases to obtain explicit knowledge or utilized large language models (LLMs) to get implicit knowledge. However, it is a complicated pipeline to construct and retrieve these knowledge bases, which can introduce additional information unrelated to the question. In addition, LLM-based methods leverage image captions related to the question as contextual information to prompt LLMs, which suffer from insufficient contextual information and lack the key details necessary to answer the question. To address the issues, we propose a unified representation-generation-selection framework, named RGS, which first obtains multi-source captioning-fused image context representation based on multiple different question-aware image captioning models, then generates candidate answers via a two-stream in-context learning method combined with the image context representation, finally selects the best answer from the candidate answers by means of instruction tuning. To balance the trade-off between complexity and performance, we only additionally finetune a question-aware image captioning model named InstructCap and an optimal answer reasoning model named InstructJudge. Compared to most methods that rely on LLMs with over 100 billion parameters (e.g., GPT-3 - 175B), our approach leverages the 7B-parameter LLM Mistral-7B as implicit knowledge, achieving state-of-the-art performance on multiple knowledge-based VQA benchmarks and significantly outperforming several previous state-of-the-art methods.
- 30 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. PP (15209210) , 1-11
RGS: A Unified Representation-Generation-Selection Framework for Knowledge-Based Visual Question Answering
Knowledge-based visual question answering (VQA) is a task that answers questions with additional knowledge beyond the image itself. Existing methods have either retrieved external knowledge bases to obtain explicit knowledge or utilized large language models (LLMs) to get implicit knowledge. However, it is a complicated pipeline to construct and retrieve these knowledge bases, which can introduce additional information unrelated to the question. In addition, LLM-based methods leverage image captions related to the question as contextual information to prompt LLMs, which suffer from insufficient contextual information and lack the key details necessary to answer the question. To address the issues, we propose a unified representation-generation-selection framework, named RGS, which first obtains multi-source captioning-fused image context representation based on multiple different question-aware image captioning models, then generates candidate answers via a two-stream in-context learning method combined with the image context representation, finally selects the best answer from the candidate answers by means of instruction tuning. To balance the trade-off between complexity and performance, we only additionally finetune a question-aware image captioning model named InstructCap and an optimal answer reasoning model named InstructJudge. Compared to most methods that rely on LLMs with over 100 billion parameters (e.g., GPT-3 - 175B), our approach leverages the 7B-parameter LLM Mistral-7B as implicit knowledge, achieving state-of-the-art performance on multiple knowledge-based VQA benchmarks and significantly outperforming several previous state-of-the-art methods.
- 10 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62 (00189251) , 1538-1551
The Pareto-Optimal Tradeoff Between Sidelobe Suppression and Signal Strength for Golay Complementary Waveforms
The ambiguity function (AF) of traditional Golay complementary waveform is sidelobe free along zero-Doppler axis but it exhibits significant sidelobes at non-zero Doppler shifts, which pose challenges to unambiguous ranging for high-speed targets. In this paper, a joint transmit-receive multi-objective optimization framework is proposed for designing low range Doppler sidelobe and high signal strength Golay Complementary Waveforms. In the framework, the Pareto-optimal trade-off between sidelobe-to-mainlobe ratio (SMR) and signal-to-noise ratio (SNR) is achieved by constructing a Pareto multi-objective optimization problem, and a model-driven machine learning (ML) algorithm is further proposed to solve the optimization problem. By means of the proposed framework and algorithms, the range sidelobes in the Doppler interval of interest are suppressed to be extremely low with little SNR loss. Numerical results validate the range sidelobes in the Doppler interval [0, π) are suppressed to be-80.38 dB with SNR penalty of 2.8 dB. The achieved-80.38 dB sidelobe suppression is almost 63 dB lower than traditional GCP waveform. Moreover, range sidelobe suppression and SNR of the proposed Golay complementary waveforms in this paper are 5 dB and 0.3 dB better respectively than state-of-the-art low sidelobe modified Golay complementary waveforms in literature.
- 10 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Aerospace and Electronic Systems
- Vol. 62 (00189251) , 1538-1551
The Pareto-Optimal Tradeoff Between Sidelobe Suppression and Signal Strength for Golay Complementary Waveforms
The ambiguity function (AF) of traditional Golay complementary waveform is sidelobe free along zero-Doppler axis but it exhibits significant sidelobes at non-zero Doppler shifts, which pose challenges to unambiguous ranging for high-speed targets. In this paper, a joint transmit-receive multi-objective optimization framework is proposed for designing low range Doppler sidelobe and high signal strength Golay Complementary Waveforms. In the framework, the Pareto-optimal trade-off between sidelobe-to-mainlobe ratio (SMR) and signal-to-noise ratio (SNR) is achieved by constructing a Pareto multi-objective optimization problem, and a model-driven machine learning (ML) algorithm is further proposed to solve the optimization problem. By means of the proposed framework and algorithms, the range sidelobes in the Doppler interval of interest are suppressed to be extremely low with little SNR loss. Numerical results validate the range sidelobes in the Doppler interval [0, π) are suppressed to be-80.38 dB with SNR penalty of 2.8 dB. The achieved-80.38 dB sidelobe suppression is almost 63 dB lower than traditional GCP waveform. Moreover, range sidelobe suppression and SNR of the proposed Golay complementary waveforms in this paper are 5 dB and 0.3 dB better respectively than state-of-the-art low sidelobe modified Golay complementary waveforms in literature.
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3693-3701
DSACap: Enhancing Visual-Semantic Alignment with Diffusion-based Framework for Image Captioning
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 3693-3701
Open AccessDSACap: Enhancing Visual-Semantic Alignment with Diffusion-based Framework for Image Captioning
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (1) , 47-61
Foundation Model for Skeleton-Based Human Action Understanding
Human action understanding serves as a foundational pillar in the field of intelligent motion perception.Skeletons serve as a modality- and device-agnostic representation for human modeling, and skeleton-based action understanding has potential applications in humanoid robot control and interaction. However, existing works often lack the scalability and generalization required to handle diverse action understanding tasks. There is no skeleton foundation model that can be adapted to a wide range of action understanding tasks. This paper presents a Unified Skeleton-based Dense Representation Learning (USDRL) framework, which serves as a foundational model for skeleton-based human action understanding. USDRL consists of a Transformer-based Dense Spatio-Temporal Encoder (DSTE), Multi-Grained Feature Decorrelation (MG-FD), and Multi-Perspective Consistency Training (MPCT). The DSTE module adopts two parallel streams to learn temporal dynamic and spatial structure features. The MG-FD module collaboratively performs feature decorrelation across temporal, spatial, and instance domains to reduce dimensional redundancy and enhance information extraction. The MPCT module employs both multi-view and multi-modal self-supervised consistency training. The former enhances the learning of high-level semantics and mitigates the impact of low-level discrepancies, while the latter effectively facilitates the learning of informative multimodal features. We perform extensive experiments on 25 benchmarks across across 9 skeleton-based action understanding tasks, covering coarse prediction, dense prediction, and transferred prediction. Our approach significantly outperforms the current state-of-the-art methods. We hope that this work would broaden the scope of research in skeleton-based action understanding and encourage more attention to dense prediction tasks.
- 20 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 48 (1) , 47-61
Foundation Model for Skeleton-Based Human Action Understanding
Human action understanding serves as a foundational pillar in the field of intelligent motion perception.Skeletons serve as a modality- and device-agnostic representation for human modeling, and skeleton-based action understanding has potential applications in humanoid robot control and interaction. However, existing works often lack the scalability and generalization required to handle diverse action understanding tasks. There is no skeleton foundation model that can be adapted to a wide range of action understanding tasks. This paper presents a Unified Skeleton-based Dense Representation Learning (USDRL) framework, which serves as a foundational model for skeleton-based human action understanding. USDRL consists of a Transformer-based Dense Spatio-Temporal Encoder (DSTE), Multi-Grained Feature Decorrelation (MG-FD), and Multi-Perspective Consistency Training (MPCT). The DSTE module adopts two parallel streams to learn temporal dynamic and spatial structure features. The MG-FD module collaboratively performs feature decorrelation across temporal, spatial, and instance domains to reduce dimensional redundancy and enhance information extraction. The MPCT module employs both multi-view and multi-modal self-supervised consistency training. The former enhances the learning of high-level semantics and mitigates the impact of low-level discrepancies, while the latter effectively facilitates the learning of informative multimodal features. We perform extensive experiments on 25 benchmarks across across 9 skeleton-based action understanding tasks, covering coarse prediction, dense prediction, and transferred prediction. Our approach significantly outperforms the current state-of-the-art methods. We hope that this work would broaden the scope of research in skeleton-based action understanding and encourage more attention to dense prediction tasks.
- Open Access
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 312-320
PTSR: A Unified Patch Tokenization, Selection and Representation Framework for Efficient Micro-expression Recognition
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 312-320
Open AccessPTSR: A Unified Patch Tokenization, Selection and Representation Framework for Efficient Micro-expression Recognition
- Open Access
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1331-1339
MirrorDiff: Learning Mirror Diffusion for Image Captioning via Regeneration
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1331-1339
Open AccessMirrorDiff: Learning Mirror Diffusion for Image Captioning via Regeneration
- 8 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Vehicles
- Vol. 10 (4) , 2795-2805
Unimodular Complete Complementary Sequence With Optimal Trade-Off Between Auto- and Cross-Ambiguity Functions for MIMO Radars
It is difficult to design unimodular sequences with both good auto-ambiguity functions (AF) and cross-ambiguity functions (CAF), due to its nonconvex and higher cardinality property. To address this problem, we propose a Pareto optimization framework to reach an optimal tradeoff between AF and CAF. Besides, we also propose a novel convolution-based multi-objective optimization algorithm to optimize the conventional metrics including peak sidelobe level (PSL) and integrated sidelobe level (ISL). In this way, the generated sequences are Doppler resilient with the desired shape of aperiodic or periodic AFs. Compared with the state-of-the-art methods, simulation experimental results indicate the PSL metric of AF and CAF can be both suppressed to -72dB under the same system parameters, obtaining a minimum of 7 dB gain.
- 8 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Vehicles
- Vol. 10 (4) , 2795-2805
Unimodular Complete Complementary Sequence With Optimal Trade-Off Between Auto- and Cross-Ambiguity Functions for MIMO Radars
It is difficult to design unimodular sequences with both good auto-ambiguity functions (AF) and cross-ambiguity functions (CAF), due to its nonconvex and higher cardinality property. To address this problem, we propose a Pareto optimization framework to reach an optimal tradeoff between AF and CAF. Besides, we also propose a novel convolution-based multi-objective optimization algorithm to optimize the conventional metrics including peak sidelobe level (PSL) and integrated sidelobe level (ISL). In this way, the generated sequences are Doppler resilient with the desired shape of aperiodic or periodic AFs. Compared with the state-of-the-art methods, simulation experimental results indicate the PSL metric of AF and CAF can be both suppressed to -72dB under the same system parameters, obtaining a minimum of 7 dB gain.