Dingyuan Zhang
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
Dingyuan Zhang has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Dingkang Liang
Huazhong University of Science and Technology
4 shared publications - Xiang Bai
Huazhong University of Science and Technology, Wuhan, China
4 shared publications - Cheng Zhang
Huazhong University of Science and Technology
2 shared publications - Dewen Dong
Changchun Institute of Applied Chemistry, Chinese Academy of Sciences
2 shared publications - Jingdong Wang
Baidu, Beijing, China
2 shared publications - Ning Zhang
Faculty of Chemistry Northeast Normal University Changchun 130024 P. R. China
2 shared publications
Latest Publications
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 35, 8558-8573
Extending Large Vision–Language Model for Diverse Interactive Tasks in Autonomous Driving
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Image Processing
- Vol. 35, 8558-8573
Extending Large Vision–Language Model for Diverse Interactive Tasks in Autonomous Driving
- 23 November 2024
- Published by Springer Nature
- p. 56-72
Make Your ViT-Based Multi-view 3D Detectors Faster via Token Compression
- 23 November 2024
- Published by Springer Nature
- p. 56-72
Make Your ViT-Based Multi-view 3D Detectors Faster via Token Compression
- Open Access
Make Your ViT-based Multi-view 3D Detectors Faster via Token Compression
Slow inference speed is one of the most crucial concerns for deploying multi-view 3D detectors to tasks with high real-time requirements like autonomous driving. Although many sparse query-based methods have already attempted to improve the efficiency of 3D detectors, they neglect to consider the backbone, especially when using Vision Transformers (ViT) for better performance. To tackle this problem, we explore the efficient ViT backbones for multi-view 3D detection via token compression and propose a simple yet effective method called TokenCompression3D (ToC3D). By leveraging history object queries as foreground priors of high quality, modeling 3D motion information in them, and interacting them with image tokens through the attention mechanism, ToC3D can effectively determine the magnitude of information densities of image tokens and segment the salient foreground tokens. With the introduced dynamic router design, ToC3D can weigh more computing resources to important foreground tokens while compressing the information loss, leading to a more efficient ViT-based multi-view 3D detector. Extensive results on the large-scale nuScenes dataset show that our method can nearly maintain the performance of recent SOTA with up to 30% inference speedup, and the improvements are consistent after scaling up the ViT and input resolution. The code will be made at https://github.com/DYZhang09/ToC3D.
Open AccessMake Your ViT-based Multi-view 3D Detectors Faster via Token Compression
Slow inference speed is one of the most crucial concerns for deploying multi-view 3D detectors to tasks with high real-time requirements like autonomous driving. Although many sparse query-based methods have already attempted to improve the efficiency of 3D detectors, they neglect to consider the backbone, especially when using Vision Transformers (ViT) for better performance. To tackle this problem, we explore the efficient ViT backbones for multi-view 3D detection via token compression and propose a simple yet effective method called TokenCompression3D (ToC3D). By leveraging history object queries as foreground priors of high quality, modeling 3D motion information in them, and interacting them with image tokens through the attention mechanism, ToC3D can effectively determine the magnitude of information densities of image tokens and segment the salient foreground tokens. With the introduced dynamic router design, ToC3D can weigh more computing resources to important foreground tokens while compressing the information loss, leading to a more efficient ViT-based multi-view 3D detector. Extensive results on the large-scale nuScenes dataset show that our method can nearly maintain the performance of recent SOTA with up to 30% inference speedup, and the improvements are consistent after scaling up the ViT and input resolution. The code will be made at https://github.com/DYZhang09/ToC3D.
- 7 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 8 (11) , 7464-7471
You Only Look Bottom-Up for Monocular 3D Object Detection
Monocular 3D Object Detection is an essential task for autonomous driving. Meanwhile, accurate 3D object detection from pure images is very challenging due to the loss of depth information. Most existing image-based methods infer objects' location in 3D space based on their 2D sizes on the image plane, which usually ignores the intrinsic position clues from images, leading to unsatisfactory performances. Motivated by the fact that humans could leverage the bottom-up positional clues to locate objects in 3D space from a single image, in this paper, we explore the position modeling from the image feature column and propose a new method named You Only Look Bottum-Up (YOLOBU). Specifically, our YOLOBU leverages Column-based Cross Attention to determine how much a pixel contributes to pixels above it. Next, the Row-based Reverse Cumulative Sum (RRCS) is introduced to build the connections of pixels in the bottom-up direction. Our YOLOBU fully explores the position clues for monocular 3D detection via building the relationship of pixels from the bottom-up way. Extensive experiments on the KITTI dataset demonstrate the effectiveness and superiority of our method.
- 7 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 8 (11) , 7464-7471
You Only Look Bottom-Up for Monocular 3D Object Detection
Monocular 3D Object Detection is an essential task for autonomous driving. Meanwhile, accurate 3D object detection from pure images is very challenging due to the loss of depth information. Most existing image-based methods infer objects' location in 3D space based on their 2D sizes on the image plane, which usually ignores the intrinsic position clues from images, leading to unsatisfactory performances. Motivated by the fact that humans could leverage the bottom-up positional clues to locate objects in 3D space from a single image, in this paper, we explore the position modeling from the image feature column and propose a new method named You Only Look Bottum-Up (YOLOBU). Specifically, our YOLOBU leverages Column-based Cross Attention to determine how much a pixel contributes to pixels above it. Next, the Row-based Reverse Cumulative Sum (RRCS) is introduced to build the connections of pixels in the bottom-up direction. Our YOLOBU fully explores the position clues for monocular 3D detection via building the relationship of pixels from the bottom-up way. Extensive experiments on the KITTI dataset demonstrate the effectiveness and superiority of our method.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1029-1034
The Fault Diagnosis of Rolling Bearing Based on WPD and TPOT
In this paper, in order to tackle the current problems of rolling bearing fault diagnosis, a new kind of bearing fault diagnosis method based on Wave Packet Decomposition (WPD) and Tree-based Pipeline Optimization Tool (TPOT) is proposed. Firstly, the feature vectors of bearing fault are extracted by using the wavelet packet decomposition. Then, the genetic programming that is based on tree structures is employed to generate the optimal machine learning pipeline. The specific structure and parameters are automatically evolved to obtain the best classification performance. Finally, the method is tested on practical bearing experiments. The experimental results have shown the advantages and effectiveness of the proposed method.
- 1 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1029-1034
The Fault Diagnosis of Rolling Bearing Based on WPD and TPOT
In this paper, in order to tackle the current problems of rolling bearing fault diagnosis, a new kind of bearing fault diagnosis method based on Wave Packet Decomposition (WPD) and Tree-based Pipeline Optimization Tool (TPOT) is proposed. Firstly, the feature vectors of bearing fault are extracted by using the wavelet packet decomposition. Then, the genetic programming that is based on tree structures is employed to generate the optimal machine learning pipeline. The specific structure and parameters are automatically evolved to obtain the best classification performance. Finally, the method is tested on practical bearing experiments. The experimental results have shown the advantages and effectiveness of the proposed method.
- 26 June 2013
- Published by Royal Society of Chemistry (RSC) in Chemical Communications
- Vol. 49 (66) , 7358-7360
Formal [4+1] annulation of α,α-dialkyl β-oxo amides and dimethylsulfoxonium methylide: a synthetic route to β-hydroxy-γ-lactams
A facile and efficient one-pot synthesis of β-hydroxy-γ-lactams from α,α-dialkyl β-oxo amides and trimethylsulfoxonium iodide in the presence of sodium hydride via a tandem Corey–Chaykovsky reaction and an intramolecular lactamization process is developed.
- 26 June 2013
- Published by Royal Society of Chemistry (RSC) in Chemical Communications
- Vol. 49 (66) , 7358-7360
Formal [4+1] annulation of α,α-dialkyl β-oxo amides and dimethylsulfoxonium methylide: a synthetic route to β-hydroxy-γ-lactams
A facile and efficient one-pot synthesis of β-hydroxy-γ-lactams from α,α-dialkyl β-oxo amides and trimethylsulfoxonium iodide in the presence of sodium hydride via a tandem Corey–Chaykovsky reaction and an intramolecular lactamization process is developed.
- 3 February 2012
- Published by Georg Thieme Verlag KG in Synthesis
- Vol. 44 (05) , 705-710
Efficient and Divergent Synthesis of Functionalized Cyclopropanes via Iodoform Reaction
Efficient and divergent one-pot synthesis of cyclopropyl amides and esters from readily available 1-acetylcyclopropanes via iodoform reaction based on the selection of reaction conditions is reported. A series of substituted cyclopropyl amides were synthesized from 1-acetylcyclopropanes, iodine, and ammonia in water in the presence of K2CO3 in good yields, whereas substituted cyclopropyl esters were obtained from the reaction of 1-acetylcyclopropanes with iodine and alcohols in the presence of DBU.
- 3 February 2012
- Published by Georg Thieme Verlag KG in Synthesis
- Vol. 44 (05) , 705-710
Efficient and Divergent Synthesis of Functionalized Cyclopropanes via Iodoform Reaction
Efficient and divergent one-pot synthesis of cyclopropyl amides and esters from readily available 1-acetylcyclopropanes via iodoform reaction based on the selection of reaction conditions is reported. A series of substituted cyclopropyl amides were synthesized from 1-acetylcyclopropanes, iodine, and ammonia in water in the presence of K2CO3 in good yields, whereas substituted cyclopropyl esters were obtained from the reaction of 1-acetylcyclopropanes with iodine and alcohols in the presence of DBU.