Avishkar Saha

PhD Student @ CVSSP. Applied Scientist Intern @ Amazon (AWS).

prof_pic.jpg

I’m a final year PhD candidate at CVSSP, University of Surrey, supervised by Richard Bowden, Chris Russell and Oscar Mendez.

So far my research has explored the intersection of Computer Vision, Geometric Deep Learning and Causality. In particular, I am interested in building models which can reason about entities and their relations in the real world.

My interests include, but are not limited to: Deep Learning for Relational Reasoning, Data Efficient Machine Learning, Generative Modeling and Optimizing Neural Network Memory Consumption.

Google Scholar/Github/LinkedIn/CV

news

July 14, 2023 Our paper "Learning Adaptive Neighborhoods for Graph Neural Networks" got into ICCV 2023!
Dec 1, 2022 Started 6-month internship at Amazon AWS’ Causal Representation Learning Lab in Tubingen.
May 23, 2022 Our paper Translating Images into Maps won the Outstanding Paper Award at ICRA 2022.

selected publications

  1. sportvu.gif
    Learning Adaptive Neighborhoods for Graph Neural Networks
    Avishkar Saha, Oscar Mendez, Chris Russell, and Richard Bowden
    In 2023 International Conference on Computer Vision (ICCV), 2023
  2. message_passing.png
    "The Pedestrian next to the Lamppost" Adaptive Object Graphs for Better Instantaneous Mapping
    Avishkar Saha, Oscar Mendez, Chris Russell, and Richard Bowden
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022
  3. image_to_bev_motivation.gif
    Translating images into maps
    Avishkar Saha, Oscar Mendez, Chris Russell, and Richard Bowden
    In 2022 International Conference on Robotics and Automation (ICRA), 2022
  4. video_bev_results.gif
    Enabling spatio-temporal aggregation in birds-eye-view vehicle estimation
    Avishkar Saha, Oscar Mendez, Chris Russell, and Richard Bowden
    In 2021 IEEE International Conference on Robotics and Automation (ICRA), 2021