Ashkan Khakzar

Researcher in Machine Learning and Computer Vision

prof_pic.jpg

I’m a postdoctoral researcher in machine learning and computer vision working with Philip Torr at the University of Oxford. I am driven by my curiosity about understanding how intelligence emerges in neural networks through learning. Thus my PhD research was focused on how to interpret neural networks, and I was lucky to have this experience at the Technical University of Munich (TUM). I was inspired throughout my entire PhD journey by my supervisor, Nassir Navab. I am very thankful to Bernt Schiele for reviewing my PhD research and inspiring me with his ideas. These days I am following the same vision in the context of vision-language foundation models.

news

Sep 28, 2024 Check out our ECCV 2024 workshop: Emergent Visual Abilities and Limits of Foundation Models.
Sep 27, 2024 We have a paper in NeurIPS 24 on evaluating abstract shape recognition in vision-language models (uploading soon).
Sep 12, 2024 Was awarded a grant by the Google Gemma 2 Academic Program to do research on GemmaScope
Aug 12, 2024 We have a perspective paper on the cognitive revolution in interpreting neural networks.
Jul 4, 2024 Check out our ECCV 2024 paper on safe text to image generation.
Jun 6, 2024 Check out our paper on guiding the attention of vision transformers.
Apr 25, 2024 Invited speaker at Trustworthy Multimodal Learning with Foundation Models at British Machine Vision Assosication.
Talk title: Understanding Foundation Models through Interpretation and Evaluation.

>Favorite Publications

  1. CVPR
    Do explanations explain? model knows best
    Ashkan Khakzar, Pedram Khorsandi, Rozhin Nobahari, and 1 more author
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022
  2. CVPR
    Neural Response Interpretation through the Lens of Critical Pathways
    Ashkan Khakzar, Soroosh Baselizadeh, Saurabh Khanduja, and 3 more authors
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021
  3. NeurIPS
    Fine-grained neural network explanation by identifying input features with predictive information
    Yang Zhang*, Ashkan Khakzar*, Yawei Li, and 3 more authors
    Advances in Neural Information Processing Systems, 2021
  4. MICCAI
    Towards Semantic Interpretation of Thoracic Disease and COVID-19 Diagnosis Models
    Ashkan Khakzar, Sabrina Musatian, Jonas Buchberger, and 5 more authors
    In Medical Image Computing and Computer Assisted Intervention – MICCAI 2021, 2021