Tahira Kazimi
Virginia Tech
tahirakazimi@vt.edu
Bio
Tahira Kazimi is a 2nd year Ph.D. student in Computer Science at Virginia Tech, specializing in Generative AI and Computer Vision. She is advised by Prof. Pinar Yanardag. Her research focuses on interpretability and controllability of diffusion models, with applications in classifier explainability, visual storytelling, and video generation. She is the first author of a CVPR 2025 paper introducing a training-free framework for explaining classifiers through hierarchical semantics in text-to-image diffusion. She has advanced video diffusion pipelines, designing multi-agent frameworks to enhance generation quality. Previously, she contributed to projects on 3D reasoning with NeRFs, solar energy forecasting with transformers, and medical image segmentation. Beyond research, Tahira has served as a reviewer for AAAI conferences and CVPR workshops and co-organized the ICCV Personalization in Generative AI (P13N) Workshop.
Areas of Research
- Artificial Intelligence
Explaining in Diffusion: Explaining a Classifier with Diffusion Semantics
Classifiers are important components in many computer vision tasks, serving as the foundational backbone of a wide variety of models employed across diverse applications. However, understanding the decision-making process of classifiers remains a significant challenge. We propose DiffEx, a novel method that leverages the capabilities of text-to-image diffusion models to explain classifier decisions. Unlike traditional GAN-based explainability models, which are limited to simple, single-concept analyses and typically require training a new model for each classifier, our approach can explain classifiers that focus on single concepts (such as faces or animals) as well as those that handle complex scenes involving multiple concepts. DiffEx employs vision-language models to create a hierarchical list of semantics, allowing users to identify not only the overarching semantic influences on classifiers (e.g., the `beard’ semantic in a facial classifier) but also their sub-types, such as `goatee’ or `Balbo’ beard. Our experiments demonstrate that DiffEx is able to cover a significantly broader spectrum of semantics compared to its GAN counterparts, providing a hierarchical tool that delivers a more detailed and fine-grained understanding of classifier decisions.