About Us
Our mission is to develop state-of-the-art algorithms to address real-world challenges, involving improving imaging and vision system outputs by removing degradations in raw data obtained.
Research Areas
Learning-based Iterative Algorithms
Current literature often lacks convergence guarantees, limiting trust in the solutions provided by these methods. The research goal is to develop provably convergent techniques for image restoration, ensuring reliable and certifiable results.
Deepfake Video Detection
Existing detection algorithms fail to generalize to advanced deepfake techniques, particularly those leveraging diffusion models. The objective is to design robust detection algorithms that can generalize across all deepfake video generation methods.
Video Stabilization for Fast-Moving Videos
Current algorithms struggle to stabilize videos with fast motions like running or rapid panning. The aim is to bridge this gap by developing stabilization techniques tailored for such high-motion scenarios.
Real-time AI for Video Restoration
Literature often focuses on performance over efficiency, making it difficult to deploy on edge devices. The goal is to design lightweight, efficient algorithms for real-time video restoration on resource-constrained devices.
Generative Models
Existing generative models for image and video restoration have limitations in achieving high-quality visual aesthetics or face significant computational resource requirements. We focus on creating deployable and effective generative models for real-world applications.
Our Team
PhD Students
- Annadanam Tiruvengala Sreedeepthi : Jan 2025 --
- Wanmathy (Co-guide with Prof. Saurav Prakash): Jan 2025 --
UG Students
- Tharun Anand : Oct 2024 --
- Nikhet : Sept 2024 --
- Tanish Chudiwal : Sept 2024 --
Research Intern
- Shubhi Sushil Shukla, VIT, Vellore : Jan 2025 --