Tejas Gokhale

Tejas Gokhale

Incoming Assistant Professor
Computer Science
University of Maryland, Baltimore County

Quick Links
Meyerhoff Scholars
Center for Women in Technology
Undergraduate Research Opportunities
UMBC Stats & Rankings


Office Hours

Third-Person Bio

I am a computer vision researcher working towards the design of robust and reliable systems that can understand the visual world. My research draws inspiration from principles of perception, communication, learning, and reasoning. The two focii of my lab are to address reliability from the perspective of machine learning and from the perspective of use-inspired and human-centered computing.

I received my Ph.D. from Arizona State University where I was advised by Yezhou Yang and Chitta Baral, M.S. from Carnegie Mellon University where I worked with Aswin Sankaranarayanan, and B.E. from Birla Institute of Technology and Science. During my graduate studies I worked with wonderful collaborators at Lawrence Livermore National Laboratory, Microsoft Research, and Snap Research.

Join the Group! I am recruiting BS/MS/PhD researchers to join me at UMBC in Fall 2023. Please use this form to apply! See FAQ for more.

Jun 2023 Organizing O-DRUM 2023 (Workshop on Open-Domain Reasoning Under Multi-Modal Settings) at CVPR
Apr 2023 Defended my Ph.D !!!
Feb-Apr 2023 Invited Talks on "Reliable Semantic Vision" at
  • Rochester Institute of Technology
  • SUNY Binghamton
  • Indiana University
  • University of Maryland Baltimore County
  • Case Western Reserve University
  • Colorado School of Mines
  • Temple University
Jan 2023 Delivered SERUM (Tutorial on Semantic Data Engineering under Multimodal Settings) at WACV 2023
Oct 2022 Recognized as Top Reviewer for NeurIPS 2022
Jun 2022 Organized O-DRUM 2022 (1st Workshop on Open-Domain Retrieval Under Multi-Modal Settings) at CVPR


* indicates equal contribution

End-to-end Knowledge Retrieval with Multi-modal Queries
ACL 2023
Man Luo, Zhiyuan Fang, Tejas Gokhale, Yezhou Yang, Chitta Baral

Knowledge retrieval with multi-modal queries, i.e., queries containing information split across image and text inputs, a challenging task that differs from previous work on cross-modal retrieval. A new dataset called ReMuQ, a new pretraining task for learning knowledge retrieval with multimodal queries, and a retriever model "ReViz" that can directly process input text and images to retrieve relevant knowledge in an end-to-end fashion without being dependent on intermediate modules such as object detectors or caption generators.

Mole Recruitment: Poisoning of Image Classifiers via Selective Batch Sampling
Ethan Wisdom, Tejas Gokhale, Yezhou Yang
pdf code

A data poisoning attack that confounds ML models without any manipulation of the image or label, achieved by simply leveraging the most confounding natural samples found within the training data itself. We show the efficacy of this novel attack in offline as well as continual learning (CL) settings in image classification, thereby exposing a previously undetected vulnerability of image classifiers.

Benchmarking Spatial Relationships in Text-to-Image Generation
Tejas Gokhale, Hamid Palangi, Besmira Nushi, Vibhav Vineet, Eric Horvitz, Ece Kamar, Chitta Baral, Yezhou Yang
pdf web

We report a surprising finding that, although recent state-of-the-art T2I models exhibit high image quality, they are severely limited in their ability to generate multiple objects or the specified spatial relations such as left/right/above/below. We introduce a metric called VISOR to quantify spatial reasoning performance. VISOR can be used off-the-shelf with any text-to-image model. We construct and make available SR2D, a dataset which contains sentences that describe spatial relationships (left/right/above/below) between a pair of commonly occurring objects.

Improving Diversity with Adversarially Learned Transformations for Domain Generalization
WACV 2023
Tejas Gokhale, Rushil Anirudh, Jayaraman Thiagarajan, Bhavya Kailkhura, Chitta Baral, Yezhou Yang
pdf code video

ALT discovers diverse and adversarial transformations using an image-to-image neural network with learnable weights. ALT improves the state-of-the-art single domain generalization performance on three benchmarks and is significantly better than pixel-wise adversarial training and standard data augmentation techniques.

CRIPP-VQA: Counterfactual Reasoning about Implicit Physical Properties via Video Question Answering
EMNLP 2022
Maitreya Patel, Tejas Gokhale, Chitta Baral, Yezhou Yang
pdf web code

Although the imaging pipeline is unable to capture many physical properties of objects (eg. mass and coefficient of friction), these properties can be estimated by utilizing cues introduced by collisions. We introduce a new dataset (CRIPP-VQA) for reasoning about the implicit physical properties of objects from videos. The dataset contains videos of objects in motion, annotated with hypothetical/counterfactual questions about the effect of actions (removing/adding/replacing objects) and questions about planning (performing actions to reach a goal).

Covariate Shift Detection via Domain Interpolation Sensitivity
[SPOTLIGHT!] NeurIPS 2022 Workshop on Interpolation and Beyond
Tejas Gokhale, Joshua Feinglass, Yezhou Yang
pdf video

In this paper, we introduce a benchmark for covariate shift detection (CSD), that builds upon and complements previous work on domain generalization. We find that existing novelty detection methods designed for OOD benchmarks perform worse than simple confidence-based methods on our CSD benchmark. We propose Domain Interpolation Sensitivity (DIS), based on the simple hypothesis that interpolation between the test input and randomly sampled inputs from the training domain, offers sufficient information to distinguish between the training domain and unseen domains under covariate shift.

Unsupervised Natural Language Inference Using PHL Triplet Generation
ACL Findings 2022
Neeraj Varshney, Pratyay Banerjee, Tejas Gokhale, Chitta Baral,
pdf code video

Natural Language Inference (NLI) under three low-data settings (with missing labels; with missing labels and hypothesis; and with missing labels, hypotheses, and premises). A procedural data generation approach that leverages a set of sentence transformations to collect PHL (Premise, Hypothesis, Label) triplets for training NLI models, bypassing the need for human-annotated training data. State-of-the-art results under all three "unsupervised" settings.

Semantically Distributed Robust Optimization for Vision-and-Language Inference
ACL Findings 2022
Tejas Gokhale, Abhishek Chaudhary, Pratyay Banerjee, Chitta Baral, Yezhou Yang,
pdf code video

SDRO: a distributed robust optimization method that operates with linguistic transformations of sentence inputs, SISP: a suit of semantics-inverting (SI) and semantics-preserving (SP) linguistic transformations, and an ensembling technique for vision-and-language inference.

Generalized but not Robust? Comparing the Effects of Data Modification Methods on Out-of-Domain Generalization and Adversarial Robustness
ACL Findings 2022
Tejas Gokhale, Man Luo, Swaroop Mishra, Bhavdeep Singh Sachdeva, Chitta Baral
pdf video

In this work, we conduct a comprehensive study of common data modification strategies and evaluate not only their in-domain and OOD performance, but also their adversarial robustness (AR). This work serves as an empirical study towards understanding the relationship between generalizing to unseen domains and defending against adversarial perturbations.

To Find Waldo You Need Contextual Cues: Debiasing Who's Waldo
ACL 2022
Yiran Luo, Pratyay Banerjee, Tejas Gokhale, Yezhou Yang, Chitta Baral
pdf data

We present a debiased dataset for the Person Centric Visual Grounding (PCVG) task. For instance, in many cases the first name in the sentence corresponds to the largest bounding box, or the sequence of names in the sentence corresponds to an exact left-to-right order in the image). The debiased dataset offers the PCVG task a more practical baseline for reliable benchmarking and future improvements.

Improving Biomedical Information Retrieval with Neural Retrievers
AAAI 2022
Man Luo, Arindam Mitra , Tejas Gokhale, Chitta Baral

We seek to improve information retrieval (IR) using neural retrievers (NR) in the biomedical domain, using a three-pronged approach. (1) a template-based question generation method, (2) two novel pre-training tasks that are closely aligned to the downstream task of information retrieval, (3) the ``Poly-DPR'' model which encodes each context into multiple context vectors.

Weakly Supervised Relative Spatial Reasoning for Visual Question Answering
ICCV 2021
Pratyay Banerjee, Tejas Gokhale, Yezhou Yang, Chitta Baral

VQA models trained with two additional objectives: object centroid estimation and relative position estimation, lead to improved performance on spatial reasoning questions (in GQA) in fully supervised and few shot settings as well as improved O.O.D. generalization.

WeaQA: Weak Supervision via Captions for Visual Question Answering
ACL 2021 Findings
Pratyay Banerjee, Tejas Gokhale, Yezhou Yang, Chitta Baral

We show that models can be trained without any human-annotated Q-A pairs, but only with images and associated text captions. Our experiments suggest gains on benchmark with shifted priors (VQA-CP) over baselines which use full supervision from human-authored QA data.

HalluciNet: Scene Completion by Exploiting Object Co-occurrence Relationships
CVPR 2021 Workshop, "AI for Content Creation"
Kuldeep Kulkarni, Tejas Gokhale, Rajhans Singh, Pavan Turaga, Aswin Sankaranarayanan

Scene completion from sparse and incomplete label maps. `Halluci-Net' is a 2-stage method that captures the object co-occurrence relationships, to produce dense label maps from incomplete labelmaps and object boundaries, for image synthesis.

Self-Supervised Test-Time Learning for Reading Comprehension
NAACL 2021
Pratyay Banerjee, Tejas Gokhale, Chitta Baral

Unsupervised Reading Comprehension method that operates directly on a single test passage. Synthetic QA pairs are generated from the passage, and models are trained on these. When a new human-authored test question appears, models infer answers better than previous unsupervised methods.

Attribute-Guided Adversarial Training for Robustness to Natural Perturbations
AAAI 2021
Tejas Gokhale, Rushil Anirudh, Bhavya Kailkhura, Jayaraman Thiagarajan, Chitta Baral, Yezhou Yang
pdf code

An adversarial training approach which learns to generate new samples so as to maximize exposure of the classifier to attributes-space. Studies robustness to semantic shifts that are beyond L-p norm perturbations, on 3 types of naturally occurring perturbations --- object-related shifts, geometric transformations, and common image corruptions.

MUTANT: A Training Paradigm for Out-of-Distribution Generalization in Visual Question Answering
EMNLP 2020
Tejas Gokhale*, Pratyay Banerjee*, Chitta Baral, Yezhou Yang,
pdf data

MUTANT is a training paradigm that exposes VQA models to perceptually similar, yet semantically distinct mutations of the input image or question. We use a pairwise consistency loss between answers to original and mutant inputs as a regularization, along with an answer embedding NCE loss. MUTANT establishes a new SOTA (+10%) on the VQA-CP challenge (for generalization under Changing Priors)

Video2Commonsense: Generating Commonsense Descriptions to Enrich Video Captioning
EMNLP 2020
Zhiyuan Fang* Tejas Gokhale*, Pratyay Banerjee, Chitta Baral, Yezhou Yang,
pdf code web

Actions in videos are inherently linked to latent social and commonsense aspects. We present the first work on generating commonsense captions directly from videos, to describe latent intentions, attributes, and effects of humans in videos. Additionally we explore the use of open-ended video-based commonsense question answering (V2C-QA) as a way to enrich our captions.

VQA-LOL: Visual Question Answering under the Lens of Logic
ECCV 2020
Tejas Gokhale*, Pratyay Banerjee*, Chitta Baral, Yezhou Yang,
pdf, web video

VQA models struggle at negation, antonyms, conjunction, disjunction! We show a capability of answering logically composed questions with our novel modules and datasets, while retaining performance on VQA data.

Cooking With Blocks : A Recipe for Visual Reasoning on Image-Pairs
CVPR 2019 Workshop, Vision Meets Cognition
Tejas Gokhale, Shailaja Sampat, Zhiyuan Fang, Yezhou Yang, Chitta Baral,
pdf, [CVPR-VMC Paper] web

Given two images (source, target) with different object configurations, what is the sequence of steps to re-arrange source to match target? For this reasoning task, our modular approach that contains a visual encoder and an event-sequencer/planner, and exhibits inductive generalization.

Website theme inspirations: Alane Suhr, Stephen MacNeil, Jon Barron