Sanchar Palit

Sanchar Palit

Ph.D. Research Scholar in 3D Computer Vision, Low-Level Vision—particularly Image Super-Resolution and Video Frame Interpolation, Stochastic Modeling in Computer Vision, Event-based Vision, and Bayesian Learning.

Posts by Collection

portfolio

publications

Prototypical Quadruplet for Few-Shot Class Incremental Learning

Published in Procedia Computer Science, 2023

A prototypical few-shot class-incremental learning framework with contrastive embedding optimization and prototype re-centering for mitigating catastrophic forgetting and improving robust classification under limited data settings

Recommended citation: Sanchar Palit, Subhasis Chaudhuri, and Biplab Banerjee, "Prototypical Quadruplet for Few-Shot Class Incremental Learning", Procedia Computer Science, presented at INNS DLIA, IJCNN 2023
Download Paper | Download Slides

Variational Distribution and Experience Replay for 3D Reconstruction in a Continual Learning Framework

Published in Proceedings of the Fifteenth Indian Conference on Computer Vision Graphics and Image Processing, 2024

A continual learning framework for single-image 3D reconstruction that leverages variational latent priors and saliency map-based replay to preserve geometric structure and accurately reconstruct both previously seen and newly learned object classes. Best Paper Award — ICVGIP 2024

Recommended citation: Sanchar Palit and Sandika Biswas, "Variational Distribution and Experience Replay for 3D Reconstruction in a Continual Learning Framework", ICVGIP, 2024.
Download Paper

Revised Regularization for Efficient Continual Learning through Correlation-Based Parameter Update in Bayesian Neural Networks

Published in Proceedings of the Fifteenth Indian Conference on Computer Vision Graphics and Image Processing, 2024

A variational continual learning framework for Bayesian neural networks with parameter-efficient distribution learning, revised KL-based regularization, and importance-weighted ELBO for reducing catastrophic forgetting and improving knowledge transfer

Recommended citation: Sanchar Palit, Subhasis Chaudhuri, and Biplab Banerjee, "Revised Regularization for Efficient Continual Learning through Correlation-Based Parameter Update in Bayesian Neural Networks", ICVGIP, 2024.
Download Paper

Local-global context-aware and structure-preserving image super-resolution

Published in Neurocomputing, 2026

A multi-step diffusion-based image super-resolution framework that integrates local-global context-aware attention and distribution alignment for perceptually faithful image reconstruction.

Recommended citation: Sanchar Palit, Subhasis Chaudhuri, and Biplab Banerjee, "Local-Global Context-Aware and Structure-Preserving Image Super-Resolution," Neurocomputing, Elsevier, 2026.
Download Paper | Download Slides

Discontinuity-Preserving Image Super-Resolution using MRF-Regularized One-Step Diffusion

Published in Transactions on Machine Learning Research, 2026

A single-step diffusion-based image super-resolution framework with MRF prior regularization for efficient, sharp, and fast image super-resolution.

Recommended citation: Sanchar Palit, Subhasis Chaudhuri, and Biplab Banerjee, "Discontinuity-Preserving Image Super-Resolution using MRF-Regularized One-Step Diffusion", accepted at Transactions on Machine Learning Research (TMLR), 2026
Download Paper | Download Slides | Download Bibtex

talks

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.