Skip Navigation

Sathyanarayanan N. Aakur, Ph.D.

Event perception, Computer Vision, Abductive reasoning for problem solving, Genome/Metagenome analysis

Degree: PhD, Computer Science
Broad Expertise/Interests:
  • Computer Vision
  • Artificial Intelligence
  • Metagenome analysis
Equipment/Techniques:
  • High performance computing clusters
  • Deep Learning for Computer Vision and Genomics
  • Neuro-symbolic reasoning for commonsense visual understanding
Seeking Collaborators with Skill Sets:
  • Neuroscientists
  • Cognitive scientists working on event perception and sensori-motor control
Neuroscience Research Interests: My research is currently focused on building computer vision systems for open-world understanding i.e., without the need for annotated training data. I take inspiration from cognitive theories of event perception to build generic event understanding models in computer vision using deep learning and neuro-symbolic reasoning. I am interested in extending this research into building active vision systems that build upon theories of sensori-motor control from neuroscience and cognitive science. I also dabble in metagenome analysis with deep learning to understand disease evolution and diagnosis.
Funding in the Last Three Years:

NSF CISE (RI), NASA

Publication Highlights:

"A perceptual prediction framework for self supervised event segmentation." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.

"Action Localization through Continual Predictive Learning." Proceedings of the IEEE/CVF European Conference on Computer Vision. 2020.

"Unsupervised Gaze Prediction in Egocentric Videos by Energy-based Surprise Modeling." arXiv preprint arXiv:2001.11580 (2020).

"Genome Sequence Classification for Animal Diagnostics with Graph Representations and Deep Neural Networks." arXiv preprint arXiv:2007.12791 (2020).

Website:

saakur.github.io