ALPINE: Improving Remote Heart Rate Estimation using Contrastive Learning
Published in Winter Conference on Applications of Computer Vision (WACV), 2023
Lokendra Birla, Sneha Shukla, Anup Kumar Gupta, Puneet Gupta
Publications
RADIANT: Better rPPG estimation using signal embeddings and Transformer
Published in Winter Conference on Applications of Computer Vision (WACV), 2023
Anup Kumar Gupta, Rupesh Kumar, Lokendra Birla, Puneet Gupta
D-NEXUS: Defending text networks using summarization
Published in Electronic Commerce Research and Applications, 2022
Anup Kumar Gupta, Aryan Rastogi, Vardhan Paliwal, Fyse Nassar, Puneet Gupta
This paper proposes a novel summarization based defence, D-NEXUS (DefeNding tEXt networks Using Summarization), against adversarial attacks on the sentiment analysis models.
Download here
TRIESTE: Translation based defense for text classifiers
Published in Journal of Ambient Intelligence and Humanized Computing, 2022
Anup Kumar Gupta, Vardhan Paliwal, Aryan Rastogi, Puneet Gupta
This paper put forward an effective transformation-based defense, TRIESTE (TRanslatIon basEd defenSe for Text classifiErs), that uses back-to-back translation for defending against adversarial attacks on Text Classifiers.
Download here
Contextual Emotion Learning Challenge
Published in IEEE International Conference on Automatic Face and Gesture Recognition (FG 2021), 2022
Jainendra Shukla, Puneet Gupta, Aniket Bera, Arka Sarkar, Prakhar Goel, Shubhangi Butta, Anup Kumar Gupta and others
This paper proposes Indian Contextual Emotion Recognition (ICER) dataset based on the multi-ethnic Indian context.
Download here
DARE: Deceiving Audio-Visual Speech Recognition model
Published in Knowledge-Based Systems, 2021
Saumya Mishra, Anup Kumar Gupta, and Puneet Gupta
This paper proposes a novel attack to breach the state-of-the-art AVSR models by simultaneously deceiving the detection algorithm, with an attack success rate of 100%.
Download here
FATALRead - Fooling visual speech recognition models
Published in Applied Intelligence, 2021
Anup Kumar Gupta, Puneet Gupta, and Esa Rahtu
This paper poposes a novel adversarial attack on the state-of-the-art VSR models with a 100% attack success rate. The algorithm was able to circumvent popular transformation defences with an attack success rate of more than 99%.
Download here