Dhwanil R. Chauhan

chauha56@purdue.edu

Hammond, Indiana

I am a graduate researcher at the Center for Innovation Through Visualization and Simulation (CIVS) at Purdue University Northwest, advised by Yang Ni. With my research, I want to understand: How does an agent know enough about its world to know what to do next? This leads me to study intelligent agents that must make sense of complex environments from multimodal, incomplete, and sometimes unreliable information, retain the context that matters, recognize when important information is missing, and decide whether to seek more information or act.

My current work approaches different parts of this problem. I have worked on geometry-grounded audio-visual reasoning, multi-camera spatial understanding for industrial safety, robustness of vision-language retrieval under joint corruption, and conversational systems that identify missing information and ask targeted follow-up questions. My current work extends that last direction toward progressive evidence-driven root cause analysis: maintaining competing causal explanations, identifying consequential information gaps, seeking distinguishing evidence, and revising conclusions as new information arrives. Much of this work grows out of real industrial environments through CIVS and the Steel Manufacturing Simulation and Visualization Consortium, where information is distributed across cameras, language, sensors, and prior context, and reliability is a requirement rather than an afterthought.

Explore my research →

I am applying to PhD programs for Fall 2027 to study how intelligent agents can build and maintain task-relevant understanding of complex environments, reason about what they know and what they still need to know, and use that understanding for information seeking, planning, and action.

Before Purdue, I completed my undergraduate studies at Charotar University of Science and Technology (CHARUSAT), India, where I worked across several areas of machine learning and developed my foundations in research.

news

Mar 31, 2026	Our paper Visual Geometry Grounded Novel-View Acoustic Synthesis has been accepted at the CVPR 2026 MMFM5 Workshop. First unified framework for novel-view acoustic synthesis bypassing explicit 3D reconstruction via feed-forward visual geometry grounding.
Feb 10, 2026	Our paper Development of Trialing Image Detection for a Melt Shop Safety Tool has been accepted at AISTech 2026. Multi-camera spatial reasoning system for real-time dynamic safety zone reconfiguration in active industrial environments.

selected publications

CVPR Workshop
Visual Geometry Grounded Novel-View Acoustic Synthesis

Jay Polra, Dhwanil Chauhan, Wenjun Huang, and 3 more authors

In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2026

Purdue University Northwest · UC Irvine · San Diego State University

Abs Bib HTML

We present the first unified framework for novel-view acoustic synthesis that entirely bypasses explicit 3D visual rendering and costly photogrammetry by directly grounding spatial audio generation in feed-forward visual geometry. Our framework synthesizes accurate and immersive spatial audio in 3D spaces without requiring viewpoint images, dense point maps, or ground-truth poses for input video. We propose the Geometry-Grounded Acoustic Decoder (GGAD) to dynamically attend to cross-modal features embedding local and global geometries in audio and visual modalities. Extensive experiments show that our framework outperforms prior work across various benchmarks in high-quality, viewpoint-accurate spatial audio synthesis, without requiring time-consuming explicit rendering of novel-view images or dense point maps.
@inproceedings{polra2026nvas, title = {Visual Geometry Grounded Novel-View Acoustic Synthesis}, author = {Polra, Jay and Chauhan, Dhwanil and Huang, Wenjun and Toth, Kyle and Wang, Xianhui and Ni, Yang}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops}, year = {2026}, url = {https://dhwanil832.github.io/projects/nvas/}, note = {Purdue University Northwest · UC Irvine · San Diego State University} }
AISTech
Development of Trialing Image Detection for a Melt Shop Safety Tool

Kyle Toth, Jay Polra, Dhwanil Chauhan, and 4 more authors

In AISTech 2026 — Iron and Steel Technology Conference Proceedings, 2026

Purdue University Northwest · CIVS · SMSVC

Abs Bib HTML

We present a multi-camera perception system for dynamic safety zone reconfiguration in active industrial environments. Static safety zone definitions fail in dynamic industrial floors where equipment and personnel positions shift continuously. Our system fuses four simultaneous camera perspectives using a rule-based spatial reasoning engine to redefine green zone boundaries in real time based on outermost detected blocker positions, enabling proactive hazard detection before incidents occur.
@inproceedings{toth2026safety, title = {Development of Trialing Image Detection for a Melt Shop Safety Tool}, author = {Toth, Kyle and Polra, Jay and Chauhan, Dhwanil and Ni, Yang and Zhou, Chenn and Fisher, Conrad and Page, Garrett}, booktitle = {AISTech 2026 — Iron and Steel Technology Conference Proceedings}, year = {2026}, url = {https://dhwanil832.github.io/publications/}, note = {Purdue University Northwest · CIVS · SMSVC} }
WACV
TandemBench: A Joint-Corruption Benchmark & Analysis of Cross-Modal Retrieval

Dhwanil Chauhan and others

2026

Submitted to WACV 2027 Datasets Track

Abs Bib

TandemBench evaluates cross-modal retrieval under simultaneous image and text degradation across 19 dual-encoder models, 16 image corruptions, 12 text perturbations, five severity levels, MS-COCO and Flickr30K, and both retrieval directions. Across 47,360 primary matched conditions, joint degradation exceeds the larger single-modality loss in 95.15 percent of cases by a mean 5.07 Recall@1 points. The study further examines when joint degradation can be reconstructed from single-modality effects and how this relationship changes as corruption severity increases.
@misc{chauhan2026tandembench, title = {{TandemBench}: A Joint-Corruption Benchmark \& Analysis of Cross-Modal Retrieval}, author = {Chauhan, Dhwanil and others}, year = {2026}, url = {https://dhwanil832.github.io/projects/}, note = {Submitted to WACV 2027 Datasets Track}, }
IMMI
A Dual-Model Approach to Industrial Safety: Computer Vision for PPE Compliance and Hazard-Zone Monitoring in Steel Production

Kyle Toth, Monika Singhal, Dhwanil Chauhan, and 4 more authors

Integrating Materials and Manufacturing Innovation, May 2026

Purdue University Northwest · CIVS · SMSVC

Abs DOI Bib HTML

Experimental evaluation demonstrates strong performance, achieving an F1 score of 0.92 for worker detection and 0.93 for personal protective equipment (PPE) classification. The integrated hazard-zone monitoring module further enables spatial safety enforcement, achieving a frame-level detection accuracy of 92% for restricted-area violations. The system supports real-time monitoring and processes video streams at approximately 35 ms per frame on a standard GPU. Preliminary deployment observations in an operational steel facility suggest that the system can assist safety personnel by improving visibility of PPE compliance and restricted-area monitoring. These results demonstrate the technical feasibility of the proposed modular framework as a scalable approach for automated industrial safety monitoring.
@article{Toth2026, author = {Toth, Kyle and Singhal, Monika and Chauhan, Dhwanil and Polra, Jay and Zhou, Chenn and Page, Garrett and Fisher, Conrad}, title = {A Dual-Model Approach to Industrial Safety: Computer Vision for PPE Compliance and Hazard-Zone Monitoring in Steel Production}, journal = {Integrating Materials and Manufacturing Innovation}, year = {2026}, month = may, day = {12}, issn = {2193-9772}, doi = {10.1007/s40192-026-00461-6}, note = {Purdue University Northwest · CIVS · SMSVC}, url = {https://doi.org/10.1007/s40192-026-00461-6} }