Research

Members of IVAL at MBZUAI conduct research primarily in Multi-modal learning, Remote sensing and earth vision, Person centric vision, Synthesis and Generation, Visual Recognition with Limited Supervision and Medical Image Analysis

Multi-modal Learning

Deep learning with multiple modalities of data including images, videos, text and audio. Our team at IVAL explores how multi-modal understanding can be used for improving generalization in vision applications such as open-vocabulary object detection, and visual question-answering.

Person Centric Vision

Under the theme of human-centric visual analysis using deep learning, our team at IVAL explores problems including face detection, facial landmark localization, pedestrian detection and abnormal activity detection.

Visual Recognition with Limited Supervision

The theme explores various generic vision applications using deep learning with limited supervision including semi-supervised, weakly-supervised, few-shot, zero-shot, and self-supervised learning.

Remote Sensing and Earth Vision

Remote sensing data provide an unbiased, uninterrupted and borderless view of human activities and natural processes. Powered by data collected from diverse satellite systems, deep-learning models can be used to gain insight to guide targeted actions in remote locations.

Synthesis and Generation

Our team at IVAL researches new methods for generating high-resolution, realistic images and videos, and generation of synthetic data. Our team also works on fundamental concepts in generative models.

Medical Image Analysis

Our team at IVAL investigates real-world healthcare problems using medical imaging with deep learning and computer vision. We are also interested in investigating fundamental concepts in medical imaging such as multi-organ segmentation, medical image generation and deep-learning architectures for medical imaging.

See Publications