Research

My research focused on training computer vision based model with limited supervision labels using novel techniques in the field of Domain adaptation, Self-supervised learning, Semi-supervised learning, Regularization etc.

View my Google Scholar Profile

Publications

Unsupervised Action Anticipation Through Action Cluster Prediction

PDF

Authors: Jiuxu Chen, Nupur Thakur, Sachin Chhabra, Baoxin Li

Venue: IEEE Open Journal of Signal Processing (OJSP), 2025

Summary: An unsupervised technique for prediction human actions.

Unsupervised Action Anticipation

Label Smoothing++: Enhanced Label Regularization for Training Neural Networks

PDF GitHub Website

Authors: Sachin Chhabra, Hemanth Venkateswara, Baoxin Li

Venue: British Machine Vision Conference (BMVC), 2024

Summary: A label regularization technique that learns optimal training targets.

Label Smoothing++ Label Regularization

PatchRot: Self-Supervised Training of Vision Transformers by Rotation Prediction

PDF GitHub Website Patent

Authors: Sachin Chhabra, Hemanth Venkateswara, Baoxin Li

Venue: British Machine Vision Conference (BMVC), 2024

Summary: Self-supervised learning approach for training vision transformers by predicting the rotation angles of randomly rotated images and patches.

PatchRot Self-Supervised Learning Vision Transformers

Translation of Partially Paired Images with Generative Adversarial Networks

PDF GitHub

Authors: Sachin Chhabra, Yaoxin Zhuo, Riti Paul, Javad Sohankar, Ji Luo, Shan Li, Wendy Lee, Yi Su, Teresa Wu, Baoxin Li

Venue: IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI), 2024

Summary: A Generative Adversarial Network for translating partially paired medical images such as MRI-PET scans by leveraging both paired and unpaired data.

PaPaGAN Partially Paired Images Image Translation GAN

Generative Alignment of Posterior Probabilities for Source-free Domain Adaptation

PDF Webpage

Authors: Sachin Chhabra, Hemanth Venkateswara, Baoxin Li

Venue: Winter Conference on Applications of Computer Vision (WACV), 2023

Summary: A source-free domain adaptation method that aligns target predictions by modeling source class-wise posterior distributions.

GAP Unsupervised Domain Adaptation Posterior Alignment

PatchSwap: A Regularization Technique for Vision Transformers

PDF GitHub Webpage Patent

Authors: Sachin Chhabra, Hemanth Venkateswara, Baoxin Li

Venue: British Machine Vision Conference (BMVC), 2022

Summary: A regularization technique for Vision Transformers that improves generalization by swapping patches between images during training.

PatchSwap Regularization Vision Transformers

A Cognitive Perspective on Subjective and Objective Diagnostic Image Quality Models

PDF

Authors: JE Caviedes, BK Patel, R Gutzwiller, B Li, R Bhat, Sachin Chhabra

Venue: International Conference on Image Processing (ICIP), 2022

Summary: A study exploring diagnostic image quality through both subjective perception and objective metrics.

Image Quality

Glocal Alignment for Unsupervised Domain Adaptation

PDF GitHub

Authors: Sachin Chhabra, Prabal Bijoy Dutta, Hemanth Venkateswara, Baoxin Li

Venue: ACM Multimedia Workshop on Multimedia Understanding with Less Labeling (MULL), 2021

Summary: An unsupervised domain adaptation approach that aligns global and local features to bridge domain gaps effectively.

Glocal Unsupervised Domain Adaptation

Iterative Image Translation for Unsupervised Domain Adaptation

PDF

Authors: Sachin Chhabra, Hemanth Venkateswara, Baoxin Li

Venue: ACM Multimedia Workshop on Multimedia Understanding with Less Labeling (MULL), 2021

Summary: An unsupervised domain adaptation method that iteratively train image generator and image classifier using each other.

Iterative Image Translation Unsupervised Domain Adaptation

LLS: Regulating Neural Network Training via Learnable Label Smoothing

PDF

Authors: Sachin Chhabra, Prasanth Sai Gouripeddi, Hemanth Venkateswara, Baoxin Li

Venue: Preprint at Arxiv

Summary: A label regularization to learn adaptively calibrate neural network training targets.

Label Regularization PrePrint

Ongoing Research

  • Self-supervised techique for Transformers

    Researching a novel Self-supervised technique for training transformers from scratch.

  • Specialized LLMs

    Exploring training LLM-style models from scratch for specialized use cases.

Professional Services

Reviewer

  • The Association for the Advancement of Artificial Intelligence (AAAI), 2026
  • British Machine Vision Conference (BMVC), 2025
  • International Conference on Machine Learning (ICML), 2025
  • International Conference on Learning Representations (ICLR), 2025
  • Transactions on Intelligent Systems and Technology (TIST), 2025 View on Web of Science
  • Neural Information Processing Systems (NeurIPS), 2024
  • Outstanding Reviewer for British Machine Vision Conference (BMVC), 2024 View Recognition
  • Transactions on Intelligent Systems and Technology (TIST), 2024 View on Web of Science
  • International Conference on Computer Vision (ICCV), 2023
  • Neural Information Processing Systems (NeurIPS), 2023
  • British Machine Vision Conference (BMVC), 2023
  • Transactions on Intelligent Systems and Technology (TIST), 2023