Workflow Chair
Association for the Advancement of Artificial Intelligence (AAAI)
Handled workflow responsibilities for the AAAI 2024 Conference in Vancouver, Canada.
Curriculum vitae · Updated September 2026
Bhavan Vasu
Ph.D. researcher in machine learning & computer vision, Oregon State University
Oregon, USAvasub@oregonstate.eduLinkedInGitHubGoogle Scholar
01
Experience
Graduate Research Fellow
NASA Jet Propulsion Laboratory (JPL), Caltech · Machine Learning and Instrument Autonomy (MLIA) Group
Interpretable ML for the Planetary Data System under Steven Lu and Dr. Kiri Wagstaff.
Graduate Assistant
Oregon State University
Research interest: global explanations for complex human activity data.
Research and Development Engineer
Kitware Inc. · Explainable AI, remote sensing, interactive machine learning and retrieval
- Software developer on the Explainable AI Toolkit, under Dr. Brian Hu
- Researcher on AFRL DIY-AI, under Dr. Brian Clipp and Daniel Davila
- Software development on DARPA Explainable AI, under Dr. Anthony Hoogs
- Software development lead on US-SOCOM Human-Machine Teaming, under Dr. Arslan Basharat
- Research engineer on AFRL VIGILANT, under Dr. Rusty Blue and William Hicks
Research and Development Intern
Kitware Inc. · Image generation, domain adaptation
Research intern on AFRL VIGILANT, under Dr. Rusty Blue and Dr. Charles Law.
Graduate Research Assistant
Rochester Institute of Technology · Change/anomaly detection, remote sensing, image translation, sensor modeling
Real-time Vision and Image Processing Lab under Prof. Andreas Savakis and Prof. John Kerekes.
02
Publications
Disentangling Where and When: Factored Spatio-Temporal Explanations for Video Action Recognition
The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS), 2026
Learning Probabilistic Logic Programs with Functional Gradient Guided Language Models
Sixth International Joint Conference on Learning and Reasoning (IJCLR), Valencia, Spain, 2026 · *Equal contribution
Local-to-Global Logical Explanations for Deep Vision Models
Fifth International Joint Conference on Learning and Reasoning (IJCLR), Guildford, UK, 2025
Beyond Local Explanations: A Framework for Global Concept-Based Interpretation in Image Classification
Electronics, 2025
Interactive Mars Image Content-Based Search with Interpretable Machine Learning
Thirty-Eighth Conference on Innovative Applications of Artificial Intelligence (IAAI), Vancouver, Canada, 2024
Global Explanations for Image Classifiers
Thirty-Seventh AAAI Conference on Artificial Intelligence, Student Abstract Program, Washington, DC, 2023
Content-based Search of Large Image Archives at PDS Imaging Node
American Geophysical Union (AGU), 2022
X-MIR: EXplainable Medical Image Retrieval
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2022
XAITK: The Explainable AI Toolkit
Applied AI Letters, 2021
Explainable, Interactive Content-Based Image Retrieval
Applied AI Letters, 2021
Resilience and Plasticity of Deep Network Interpretations for Aerial Imagery
IEEE Access, 2020
Iterative and Adaptive Sampling with Spatial Attention for Black-Box Model Explanations
IEEE Winter Conference on Applications of Computer Vision (WACV), 2020
Explainability for Content-Based Image Retrieval
Proceedings of the MSS National Symposium on Sensor and Data Fusion (NSSDF), 2019
Training Deep Networks for Patch-Based Search in Satellite Imagery
Proceedings of the MSS National Symposium on Sensor and Data Fusion (NSSDF), 2019
Visualizing the Resilience of Deep Convolutional Network Interpretations
CVPR Workshops, 2019
Siamese Network with Multi-Level Features for Patch-Based Change Detection in Satellite Imagery
IEEE Global Conference on Signal and Information Processing (GlobalSIP), Anaheim, CA, 2018
Deep Learning for Object Detection and Classification in Satellite Imagery
Proceedings of the MSS National Symposium on Sensor and Data Fusion (NSSDF), 2018
Resilience and Self-Healing of Deep Convolutional Object Detectors
25th IEEE International Conference on Image Processing (ICIP), 2018
Aerial-CAM: Salient Structures and Textures in Network Class Activation Maps of Aerial Imagery
IEEE 13th Image, Video, and Multidimensional Signal Processing Workshop (IVMSP), Zagorochoria, 2018
03
Projects
Disentangling Where and When: Factored Spatio-Temporal Explanations for Video Action Recognition
Explanations for video action recognition that factor evidence into where it appears within frames and when it matters across time, so spatial and temporal contributions can be inspected separately. With Giuseppe Raffa and Prasad Tadepalli.
Learning Probabilistic Logic Programs with Functional Gradient Guided Language Models
Uses language models to propose clauses for probabilistic logic programs, with functional-gradient signals guiding which rules are learned and how they are weighted. Joint work with Saurabh Mathur and Sahil Sidheekh (equal contribution), Farbod Tavakkoli, Kristian Kersting, Sriraam Natarajan and Prasad Tadepalli; presented in Valencia, Spain.
Global Explanations for Interpretable Action Detection and Classification
Deep explanation methods often yield subjective explanations because there is no common linguistic representation to express them. This work bridges that gap with a linguistic understanding of human activity detection and classification, and uses the generated global explanations to identify anomalous action sequences.
Glass Dome Racing
Simulating realistic graphics for racing games is challenging. The “Glass Dome” approach locally projects sky texture maps inside a transitory sphere that moves with the player while rotating on its axis to better simulate reality.
Interpretable Machine Learning for the Planetary Data System
Investigated transparent image search on the Planetary Data System for imagery from the Mars Curiosity Rover and Mars Reconnaissance Orbiter. Built inherently explainable deep models with minimal performance drop relative to opaque convolutional networks.
XAITK: The Explainable AI Toolkit
Contributed several black-box explainers for retrieval and classification to the open-source explainable AI toolkit, funded by DARPA and JAIC to push evidence-based validation of AI models. Received an honorable mention at the 2021 PyTorch Annual Hackathon for responsible AI development.
DARPA Explainable AI
Bridging the gap between machine learning interpretation and human mental models with explainable, advisable human-in-the-loop ML systems that improve user-task performance. As software development and research lead, explored new explanation methods and evaluated their effectiveness for mental-model alignment and user-task efficiency.
USSOCOM eXplainable Human-Machine Teaming
As research lead on xHMT, developed systems for explainable, interactive query refinement on aerial full-motion video with active learning, and examined the user-task benefit of explanations in image retrieval.
AFRL VIGILANT: Patch-Based Search for Large-Scale Geospatial Retrieval
Developed algorithms for real-time, human-in-the-loop fine-grained and attribute-based retrieval on geo-spatialized satellite imagery. Built hybrid systems combining few-shot and active learning to perform classification and retrieval simultaneously, improving performance by almost 3× over standard metric learning.
Master’s Thesis: Visualizing Resiliency of Deep Convolutional Network Interpretations for Aerial Imagery
Visualized and quantified the resiliency of popular architectures for aerial scene classification with a p-map score that evaluates class activation maps under varying degrees of failure, and mapped the networks’ ability to self-heal by retraining healthy portions after partial damage. Read the thesis ↗
AFRL VIGILANT: SimGAN for DIRSIG-Generated Aerial Imagery
Improved the accuracy of networks trained on synthetic DIRSIG (Digital Imaging and Remote Sensing Image Generation) data from 59% to 85% for classification and segmentation, implemented in Keras with a TensorFlow backend.
AFRL VIGILANT: Change Detection with Siamese Networks and Feature Selection
Proposed a Siamese neural network that detects changes in bi-temporal aerial images with 96.7% accuracy, trained and tested on DIRSIG-simulated data matched to RapidEye spectral properties for structured change detection.
Weighted Priority Arbiter Using a Finite State Machine
Designed and implemented a weighted priority arbiter to resolve resource contention between IPs such as processor cores, memories and on-chip networks, reducing allocation delay by 9%.
04
Education
Doctor of Philosophy, Electrical Engineering and Computer Science
Oregon State University, Oregon, USA · GPA 3.74 / 4.0
Master of Science in Computer Engineering
Rochester Institute of Technology, Rochester, USA · GPA 3.40 / 4.0
Bachelor of Engineering in Electronics and Communication Engineering
Global Academy of Technology, Bangalore, India · GPA 3.0 / 4.0
05
Service, awards & leadership
Program committees
NeurIPS 2026
The Fortieth Annual Conference on Neural Information Processing Systems
ICLR 2027
The Fifteenth International Conference on Learning Representations
AAAI 2024, 2025, 2026
Association for the Advancement of Artificial Intelligence
Computer Vision and Image Understanding
Ranked 69 of 273 in Engineering, Electrical & Electronic
Awards
1st place, best poster at AI Week
Oregon State University
“Most Impactful Research”
NSF Pervasive Personalized Intelligence Fall 2022 Meeting, Portland, OR
Honorable mention, PyTorch Annual Hackathon
Responsible AI development with XAITK
1st place, on-spot photography
SJBIT inter-collegiate competition, Bangalore
1st place, national abacus hunt
BRAINOBRAIN, Madras, India
Leadership
Workflow Chair, AAAI 2024
Association for the Advancement of Artificial Intelligence
Vice President, EECS Graduate Student Association
Oregon State University
Public Relations Officer, EECS Graduate Student Association
Oregon State University
Committee member, Kitware AI team
Kitware Inc.