Ting-Bing Xu (许庭兵)
Senior Computer Vision Researcher, SenseTime
Postdoctoral Fellow of "Zhuoyue" Program, BUAA
Ph.D., National Laboratory of Pattern Recognition, CASIA
Email: tingbing_xu@buaa.edu.cn
Email: tingbing_xu@buaa.edu.cn
I am a senior computer vision researcher in the Research Center in SenseTime, where I worked on the Retrieval-Augmented Generation (RAG based on Rewriter/Intent/Embedding/ Rerank/Chat Large-scale Language Models), Online/Offline Knowlege-base Retrieval, LLM post-training (SFT + RLHF), and some CV perception algorithms/applications including 2D/3D lane/roadside detection/calssification, fine-grained classification of vehicles, perception of lamp status.
From Jan. 2020 to Mar. 2022, I worked as a Postdoctoral Fellow of "Zhuoyue" Program with Prof. Guangjun Zhang (academician of chinese academy of engineering) and Prof. Zhenzhong Wei in School of Instrumentation Science and Optoelectronic Engineering, Beihang University, where I focused on the real-time aircraft detection, tracking, 6D pose estimation (0~50km distance) and knowledge distillation of consistent feature representation in the dynamic visual measurement and engineering applications of aeronautics and astronautics field.
From Sep. 2014 to Jan. 2020, I pursued my Ph.D. degree in computer applied technology in Pattern Analysis and Learning (PAL) Group at the National Labortary of Pattern Recognition (NLPR) in the Institute of Automation, Chinese Academy of Sciences (CASIA), Beijing, China, under the supervision of Prof. Cheng-Lin Liu (fellow of CAA, CAAI, IAPR and IEEE), Assoc. Prof. Xu-Yao Zhang and Assoc. Prof. Pei-Pei Yang. My research interests include methods of lightweight deep neural networks, such as binarized weight networks, lightweight network architecture, self-distillation, and dynamical channel pruning.
I was a visiting researcher with the Department of Computer Science and Intelligent Systems, Osaka Prefecture University, Osaka, Japan, in Dec. 2018, under the supervision of Prof. Koichi Kise and Assoc. Prof. Masakazu Iwamura, where I worked on the generalization ability of deep neural network include the regularization of dropout-connect/ShakeDrop, data augmentation of label smooth/mixup/triplet loss, and adversarial learning.
2024.05: 🎉🎉 One paper shared template representation for multi-object pose estimation is accepted by NN.
2023.06: 🎉🎉 One paper on Normal vector guided 6D pose estimation is accepted by TCSVT.
2023.04: 🎉🎉 I was invited as a guest editor for the special issue "Sustainable Autonomous Driving System" on Sustainability.
2023.04: 🎉🎉 One paper on weight-sharing knowledge distillation is accepted by TCSVT.
2023.02: 🎉🎉 One paper on 3d attitude measurement of aircraft landing is accepted by TIM.
2022.10: 🎉🎉 One paper on background-aware siamese tracking is accepted by IJMLC.
2022.10: 🎉🎉 One paper on pre-locate net for object detection is accepted by CJA.
2022.07: 🎉🎉 One paper on IoU-aware matching-adaptive siamese tracking is accepted by Neurocomputing.
2022.04: 🎉🎉 One paper on Rotation Group Equivariant Convolutions is accepted by MIR.
2021.10: 🎉🎉 One paper on learning complementary siamese tracking is accepted by JVCIR.
2021.08: 🎉🎉 One paper on shaping neural architectures progressively is accepted by PR.
2021.02: 🎉🎉 One paper on 6D aircraft pose from keypoints and structures is accepted by RS.
2020.12: 🎉🎉 Program Committee member invitation for IJCAI (The 30th International Joint Conference on Artificial Intelligence) 2021.
2020.10: 🎉🎉 One paper on network self-distillation of data representation invariance is accepted by TNNLS.
2020.04: 🎉🎉 One paper on dynamical channel pruning by conditional accuracy change is accepted by TNNLS.
2019.07: 🎉🎉 One paper on data-distortion guided self-distillation is accepted by AAAI.
2019.04: 🎉🎉 One paper on LightweightNet via architecture distillation is accepted by PR.