Ankit Singh

I am currently working as a Research engineer where my focus is to develop multi-modal large language models (LLMs). Previously, I was a research student at Computer Science & Engineering Department, Indian Institute of Technology, Madras ( IIT Madras ) , where I worked on Computer Vision and Deep Learning.
I did my undergraduate studies in Computer Science at National Institute of Technology, Silchar( NIT Silchar )

Email  /  Twitter  /  LinkedIn  

My research interests mainly lie in the areas of computer vision and deep learning. In partiuclar, my current work is particularly focused on vision language models and label-efficient (Semi-Supervised/ Unsupervised /Self-Supervised ) approaches for deep-learning across Images/Videos.
In addition, I am also interested in video understanding, representation learning ,domain adaptation and transfer learning.

Google Scholar  /  Github

• Paper on Harnessing Frozen Unimodal Encoders for Flexible Multimodal Alignment CVPR 2025
  
• Paper on permutation symmetries in Bayesian neural network posteriors at NeurIPS 2023
  
• Institute Research Award 2022, IIT Madras
  
• Student Volunteer for NeurIPS 2021
  
• Paper on Semi-Supervised Domain Adaptation accepted at NeurIPS 2021
  
• Paper on Semi-Supervised Action Recognition accepted at CVPR 2021.
  
• Paper on Mitigating data Imbalance accepted at ECCV-W 2020.
  

In this work, we first extend the formalism of marginalized loss barrier and solution interpolation to BNNs, before proposing a matching algorithm to search for linearly connected solutions. This is achieved by aligning the distributions of two independent approximate Bayesian solutions with respect to permutation matrice

We propose a contrastive framework for semi-supervised domain adaptation (SSDA) where we use instance alignment between unlabeled target samples and centroid alignment between source and target domains.

We propose a temporal contrastive learning framework for semi-supervised action recognition by using contrastive losses between different videos and groups of videos with similar actions.

We introduce a joint dataset repairment strategy by combining classifier with a GAN that makes up for the deficit of training examples from the minority class by producing additional examples.

• Reviewer: CVPR, ICCV, ECCV, ICLR, NeurIPS, AAAI, WACV, ACCV, BMVC, TPAMI
  

Website template from here

* denotes equal contribution