Jindong Gu

Ph.D. Candidate in University of Munich

I am a final year Ph.D. student in the University of Munich, advised by Prof. Volker Tresp. My research interests include computer vision and machine learning. Especially, I am interested in robust inference and robust learning in deep vision systems from the following perspectives:

    Robust Inference:
    • Robustness: Natural and Adversarial Robustness
    • Explainability: Model Explainability and Decision Explanation
    • Efficiency: Efficient Visual Model and Inference
    Robust Learning:
    • Data Poisoning/Backdoor Attack: Learning from Malicious Data
    • Semi-supervised Learning: Learning from Partially Labeled Data
    • Few-shot Learning: Learning from A Few Samples
    • Incremental/Continual Learning: Learning from Incremental Data

    News

    Jul. 2022 Four papers on Robustness of Vision Systems have been accepted to ECCV 2022.
    Mar. 2022 I join Torr Vision Group as Visiting Ph.D. Student at University of Oxford.
    Apr. 2021 I join Microsoft Research Asia as a Research Intern.
    Mar. 2021 One paper has been accepted as oral to CVPR 2021.
    Jan. 2021 One paper has been accepted to ICLR 2021.
    Oct. 2020 I join Tencent AI Lab as a Research Intern.

    Education

    Mar. 2022 - Aug. 2022 Visiting Ph.D. Student, Torr Vision Group in University of Oxford
    Sept. 2017 - present Ph.D. Candidate, Computer Science in University of Munich
    Apr. 2016 - Jul. 2017 Master Degree, Computer Science in University of Munich
    Sept. 2011 - Mar. 2016 Double Bachelor Degrees, University of Wuppertal and Wuhan University

    Experience

    Aug. 2022 - Nov. 2022 Research Intern at Google Brain, New York, USA
    Apr. 2021 - Feb. 2022 Research Intern at Microsoft Research Asia, Beijing, China
    Aug. 2020 - Mar. 2021 Research Intern at Tencent AI Lab, Shenzhen, China
    Sept. 2017 - Aug. 2020 Doctoral Researcher at Siemens Technology, Munich, Germany

    Selected Publications

      • Are Vision Transformers robust to Patch-wise Perturbation? [PDF]
        Jindong Gu, Volker Tresp, Yao Qin
        European Conference on Computer Vision (ECCV) , 2022
      • SegPGD: An Effective and Efficient Adversarial Attack for Evaluating and Boosting Segmentation Robustness [PDF]
        Jindong Gu, Hengshuang Zhao, Volker Tresp, Phillip Torr
        European Conference on Computer Vision (ECCV) , 2022
      • Towards Efficient Adversarial Training on Vision Transformers [PDF]
        Boxi Wu*, Jindong Gu*, Zhifeng Li, Deng Cai, Xiaofei He, Wei Liu
        European Conference on Computer Vision (ECCV) , 2022
      • Watermark Vaccine: Adversarial Attacks to Prevent Watermark Removal [PDF]
        Liu Xinwei, Jian Liu, Yang Bai, Jindong Gu, Tao Chen, Xiaojun Jia, Xiaochun Cao
        European Conference on Computer Vision (ECCV) , 2022
      • Capsule Network is Not More Robust than Convolutional Network [PDF]
        Jindong Gu, Volker Tresp, Han Hu
        IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2021
      • Effective and Efficient Vote Attack on Capsule Networks [PDF][CODE]
        Jindong Gu, Baoyuan Wu, Volker Tresp
        International Conference on Learning Representations (ICLR) , 2021
      • Interpretable Graph Capsule Networks for Object Recognition [PDF] [CODE]
        Jindong Gu
        AAAI Conference on Artificial Intelligence (AAAI) , 2021
      • Simple Distillation Baselines for Improving Small Self-supervised Models [PDF] [CODE]
        Jindong Gu, Wei Liu, Yonglong Tian
        Workshop in ICCV , 2021
      • Improving the Robustness of Capsule Networks to Image Affine Transformations [PDF]
        Jindong Gu, Volker Tresp
        IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2020
      • Search for Better Students to Learn Distilled Knowledge [PDF]
        Jindong Gu, Volker Tresp
        European Conference on Artificial Intelligence (ECAI) , 2020
      • Saliency Methods for Explaining Adversarial Attacks [PDF]
        Jindong Gu, Volker Tresp
        Human-Centric Machine Learning Workshop in NeurIPS , 2019
      • Understanding Individual Decisions of CNNs via Contrastive Backpropagation [PDF] [CODE]
        Jindong Gu, Volker Tresp
        Asian Conference on Computer Vision (ACCV), 2018

    Professional Activities

    • Conference Reviewer:
      - International Conference on Machine Learning (ICML) 2022
      - International Conference on Learning Representations (ICLR) 2022
      - Advances in Neural Information Processing Systems (NeurIPS) 2021-2022
      - European Conference on Computer Vision (ECCV) 2022
      - IEEE International Conference on Computer Vision (ICCV) 2021
      - IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2021-2022
    • Journal Reviewer:
      - Elsevier - Neurocomputing
      - Elsevier - Pattern Recognition
      - Transactions of Machine Learning Research (TMLR)
      - IEEE Transactions on Neural Networks and Learning Systems (TNNLS)
      - IEEE Transactions on Circuits and Systems for Video Technology (TCSVT)
      - IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

    Patents and Inventions

    • Verification of classification decisions in Convolutional Neural Networks,
      European Patent Office, EP3654248
    • A method for computer-implemented analysis of a classification model,
      Patent Family: EP3726433, CN111832572, US2020334489
    • Siemens Inventions: 8 AI Inventions in Siemens Technology, Germany

    Honours and Awards

    • Google Stipend
    • Chinese Government Award for Outstanding Self-financed Students Abroad
    • CVPR22 Doctoral Consortium Award
    • University of Oxford Visiting Grant