Visual Computing and Biometric Security Lab (VCBSL)
The Visual Computing and Biometric Security Lab is a research group comprising undergraduates, Masters, and Ph.D. students directed by Dr. Ajita Rattani.
The lab's mission is to conduct cutting-edge research in advanced techniques concerning Visual Computing and Biometric Security.
Visual Computing is a broad term that encompasses several areas in computer science including computer vision image and video processing, machine learning,
deep learning, and pattern recognition. Biometrics is the art of recognizing individuals based on their physiological or behavioral characteristics.
Current research activities of the research group include studying the Bias and Fairness of AI systems, Deepfake Detection, Models for On-device AI, Wearable Biometrics,
Harnessing Unlabeled Data for Domain Adaptation, Computer Vision for Natural Disaster Assessment, and AI for Healthcare and Obesity Prediction.
News
- 1/02/2025 Our paper titled "Unified Face Matching and Physical-Digital Spoofing Attack Detection" is accepted at IEEE/CVF Winter Conference on Applications of Computer Vision Workshop (WACVW)" (pdf).
- 12/01/2024 Our paper titled "Demographic Bias Mitigation at Test-time using Uncertainty Estimation and Human-Machine Partnership
" is accepted at Machine Learning with Applications, Elsevier (pdf).
- 11/06/2024 Our paper titled "Machine Learning-based sEMG Signal Classification for Hand Gesture Recognition" is accepted at IEEE International Conference on Bioinformatics and Biomedicine (BIBM)(pdf).
- 10/16/2024 Our paper titled "Social Media Authentication and Combating Deepfakes using Semi-fragile Invisible Image Watermarking" is accepted at ACM Transactions on Digital Threats: Research and Practice (pdf).
- 09/06/2024 Our paper titled "FineFACE: Fair Facial Attribute Classification Leveraging Fine-grained Features" is accepted at ICPR 2024 (pdf).
- 09/06/2024 Our paper titled "CoDeiT: Contrastive Data-efficient Transformers for Deepfake Detection" is accepted at ICPR 2024 (pdf).
