Deepak Kumar

I am doctoral candidate in Engineering and Applied science program with a concentration in Computer and Information sciences at University of Massachusetts Dartmouth. I have completed my master’s in data science from University of Massachusetts Dartmouth and my bachelor’s in computer science from SZABIST (Karachi, Pakistan).

My research focuses on computer vision, transfer learning, adversarial learning and knowledge distillation. I'm currently investigating a novel way to attack multi-modality, multi-view action recognition systems, as well as proposing a defense mechanism to protect those models from various adversarial attacks.

Currently, I am a Graduate Research Assistant at Machine Intelligence and Data Analytics (MIND) Lab at University of Massachusetts, Dartmouth. My advisor is Dr. Ming Shao .

Annoucements

• [1/2022] One regular paper on "Generative Adversarial Attack on Ensemble Clustering" got accepted in Winter Conference on Applications of Computer Vision.
• [5/2021] Started summer internship at Philips Research North America .
• [1/2021] One paper on "Collaborative Knowledge Distillation for Incomplete Multi-view Action Prediction" paper has been accepted by Image and Vision Computing-ELSEVIER.
• [8/2020] Defended PhD dissertation proposal on "Adversarial Multi-view Action Recognition".
• [7/2020] One regular Paper on "Finding Achilles' Heel: Adversarial Attack on Multi-modal Action Recognition" got accepted (acceptance rate = 27.8%) in ACM Multimedia 2020.
• [5/2020] Started summer internship at Philips Research North America .
• [5/2019] Started summer internship at Philips Research North America .
• [11/2018] Our work "Cross-view Action Recognition via Joint Dictionary & Transfer Learning" has been selected for an oral presentation at 2018 New England Computer Vision Workshop .
• [5/2018] I have been selected for feature stores of MS in Data Science of UMass Dartmouth.
• [11/2017] Our work "Cross-database Mammographic Image Analysis through unsupervised domain adaptation" has been selected for an oral presentation at 2017 New England Computer Vision Workshop .
• [11/2017] One "Cross-database Mammographic Image Analysis through unsupervised domain adaptation" paper has been accepted by the Big Data Transfer Learning workshop to be held in conjuction with IEEE International Conference on Big Data (BigData) 2017.