Hinton Research Lab

Coding Competition Organized at Department of Computer Science, Gauhati University

Mon, 27 Apr 2026 00:00:00 +0000

The Department of Computer Science, Gauhati University, successfully organized a coding competition for students of the Master’s programme and FYIMP (Five Year Integrated Master Programme). The event aimed to encourage students to strengthen their problem-solving abilities, logical reasoning, and data structures & algorithms (DSA) skills through competitive programming activities.

The competition was conducted in two rounds:

MCQ Round – Designed to test participants on fundamental concepts of programming, algorithms, aptitude, and computer science basics.
Coding Round – Focused on real-time problem solving where students implemented algorithmic solutions within a limited time.

The event witnessed enthusiastic participation from students, showcasing their technical knowledge and coding capabilities. The competition provided a healthy platform for learning, innovation, and peer interaction among aspiring programmers.

After an intense and competitive evaluation process, the winners of the competition were:

🥇 1st Place: Ved Bhandary
🥈 2nd Place: Jay Prakash Nath
🥉 3rd Place: Debojit Nath

The Department congratulates all the winners and participants for their remarkable performance and active involvement in the event. Such initiatives continue to promote a strong programming culture and motivate students to excel in competitive coding and software development.

Congratulations to Dr. Pallavi Saikia on Successful PhD Thesis Defense

Wed, 25 Mar 2026 00:00:00 +0000

We are proud to congratulate Mrs. Pallavi Saikia, a valued member of our research laboratory, on the successful defense of her PhD thesis.

Her doctoral research focused on Alzheimer’s Disease Detection using Image Processing and Computer Vision Techniques, addressing important challenges in medical image analysis and AI-assisted healthcare diagnostics. Throughout her research journey, she made significant contributions to the field through numerous research publications related to neuroimaging, deep learning, and intelligent disease detection systems.

Her work reflects a strong interdisciplinary approach combining artificial intelligence, medical imaging, and computer vision for early-stage neurological disorder analysis and diagnosis.

The entire lab extends heartfelt congratulations to Dr. Pallavi Saikia on this remarkable academic achievement and wishes her continued success in her future research and professional career.

Welcoming New PhD Scholars Neeharika Sonowal and Gaurab Khaklari to the Research Group

Sun, 22 Feb 2026 00:00:00 +0000

We are pleased to welcome Neeharika Sonowal and Gaurab Khaklari as new PhD scholars to our research group at the Department of Computer Science, Gauhati University.

Both scholars have qualified for the prestigious UGC Junior Research Fellowship (JRF) and bring valuable academic and research experience to the laboratory.

Neeharika Sonowal

Neeharika joins the lab with two years of teaching experience and research specialization in Microscopy Image Super Resolution and deep learning-based image enhancement techniques. She has contributed to the field through research publications in relevant domains and is actively interested in biomedical imaging and computational microscopy.

Gaurab Khaklari

Gaurab’s research interests include Image Processing, Deep Learning, and Assamese Handwritten Compound Character Recognition. He has prior research publication experience in the area of intelligent character recognition and pattern analysis, particularly focusing on Assamese handwritten scripts and document image analysis.

We warmly welcome both researchers to the lab and look forward to their valuable contributions to ongoing and future research projects. We wish them success in their doctoral journey and research endeavors.

Super Resolution

Sat, 07 Feb 2026 00:00:00 +0000

About This Project

This project focuses on the development of advanced Image Super Resolution (SR) techniques using deep learning and modern neural network architectures. The primary objective is to reconstruct high-resolution images from low-resolution inputs while preserving structural details, textures, and perceptual quality.

The research explores multiple AI-driven approaches including:

Convolutional Neural Networks (CNNs)
Vision Transformers (ViTs)
Mamba and State Space Models (SSM)
Attention-based architectures
GAN-based image reconstruction
Diffusion-based enhancement methods

The project aims to improve image quality in domains such as biomedical imaging, microscopy, satellite imaging, surveillance, and low-quality visual data restoration.

Research Objectives

The major objectives of this project include:

Single Image Super Resolution (SISR)
Lightweight and efficient SR architectures
Attention-guided feature enhancement
Transformer-based reconstruction models
Real-world degradation handling
Perceptual quality optimization
Explainable AI for image reconstruction

Methodology

The proposed framework includes:

Image preprocessing and degradation modeling
Feature extraction using CNN and Transformer blocks
Spatial and channel attention learning
Mamba-based sequence modeling for long-range dependencies
High-resolution image reconstruction
Perceptual and quantitative quality evaluation

The project investigates architectures such as:

SRCNN
EDSR
ESRGAN
SwinIR
Vision Transformers
VMamba and Hybrid SSM Models

Applications

Potential applications include:

Microscopy image enhancement
Medical image reconstruction
Satellite and remote sensing imagery
Surveillance and security systems
Historical image restoration
Mobile photography enhancement

Results

The project aims to achieve:

Improved PSNR and SSIM metrics
Better texture and edge preservation
Efficient lightweight deployment models
Robust real-world super resolution performance

Future Scope

Future directions include:

Foundation models for image enhancement
Real-time edge deployment
Self-supervised super resolution
Video super resolution
Multimodal image restoration frameworks
AI-assisted biomedical imaging systems

Medicinal Plant Identification using Deep Learning

Wed, 28 Jan 2026 00:00:00 +0000

About This Project

This project focuses on the development of an intelligent Medicinal Plant Identification System using advanced deep learning and computer vision techniques. The primary objective is to automatically recognize medicinal plants from leaf and plant images captured in real-world environments.

The research integrates:

Deep Learning
Computer Vision
Vision Transformers
Attention Mechanisms
Mobile AI Applications
Fine-Grained Image Classification

The system is designed to support biodiversity conservation, herbal medicine documentation, and accessible AI-driven plant identification for researchers, students, healthcare practitioners, and local communities.

Research Objectives

The major objectives of the project include:

Real-world medicinal plant recognition
Leaf-based and whole-plant identification
Robust classification under varying lighting and background conditions
Development of lightweight mobile-friendly AI models
Creation of self-curated medicinal plant datasets
Explainable AI for plant identification systems

Methodology

The proposed framework includes:

Dataset collection and annotation
Image preprocessing and augmentation
Deep feature extraction using CNN and Transformer architectures
Attention-based feature refinement
Classification and confidence prediction
Deployment through mobile and web-based applications

The project explores architectures such as:

Convolutional Neural Networks (CNNs)
Vision Transformers (ViTs)
Hybrid Attention Models
Mamba and State Space Models (SSM)

Applications

Potential applications include:

Herbal medicine documentation
Biodiversity conservation
Educational plant identification tools
Smart agriculture systems
Mobile-based field identification
Digital ethnobotany research

Results

The project aims to:

Improve medicinal plant classification accuracy
Build robust real-world datasets
Develop lightweight and scalable AI models
Enhance interpretability using explainable AI techniques

Future Scope

Future directions of the project include:

Multilingual plant information systems
Cross-domain plant recognition
Real-time mobile deployment
Integration with geospatial biodiversity mapping
Large-scale foundation models for plant intelligence

ISRO-RAC-S Sponsored Research Project Awarded on Change Detection and Change Captioning

Thu, 22 Jan 2026 00:00:00 +0000

We are delighted to announce that our Principal Investigator Dr. Sanjib Kr. Kalita, Associate Professor and Head of the Department of Computer Science, Gauhati University, has been awarded a prestigious research project funded by ISRO-RAC-S.

The project focuses on Change Detection and Change Captioning using Computer Vision and Deep Learning Techniques, aiming to develop intelligent AI-based frameworks for automated analysis and interpretation of remote sensing and satellite imagery data.

The project will involve advanced research in:

Computer Vision
Deep Learning
Remote Sensing Image Analysis
Automated Scene Understanding
AI-driven Change Captioning Systems

The Co-Principal Investigators of the project are:

Dr. Diganta Kumar Pathak, Assistant Professor, Department of Computer Science, Birangana Sati Sadhani Rajyik Vishwavidyalaya, Golaghat
Prof. Minakshi Gogoi, Head, Department of Computer Science, Girijananda Chowdhury University, Guwahati
Mr. Tushar Shukla, Sci/Eng-“SE”, Space Applications Centre, ISRO, Ahmedabad

This project marks an important milestone for the research community and is expected to contribute significantly to intelligent geospatial analysis and AI-based remote sensing applications.

The entire laboratory and department congratulate the project team and wish them great success in this research endeavor.

Unconstrained Compound Characters Recognition

Wed, 07 Jan 2026 00:00:00 +0000

About This Project

This project focuses on the development of intelligent systems for Unconstrained Compound Character Recognition using deep learning and computer vision techniques. The primary objective is to accurately recognize handwritten compound characters from unconstrained document images, particularly for Assamese and other Indic scripts.

The project addresses challenges such as:

Complex handwritten character structures
Variations in writing styles
Noise and document degradation
Character overlapping and segmentation difficulties
Low-resource language datasets

The research integrates:

Deep Learning
Optical Character Recognition (OCR)
Vision Transformers
Sequence Modeling
Attention Mechanisms
State Space Models (Mamba)

to build robust handwritten script recognition systems.

Research Objectives

The major objectives include:

Recognition of unconstrained handwritten compound characters
Robust feature extraction from document images
Lightweight and efficient OCR architectures
Handling complex ligatures and script variations
Development of annotated handwritten datasets
Improving recognition accuracy under real-world conditions

Methodology

The proposed framework includes:

Handwritten document image collection
Image preprocessing and noise removal
Character segmentation and normalization
Deep feature extraction using CNN and Transformer architectures
Sequence modeling using attention and Mamba-based networks
Compound character classification and recognition

The project investigates architectures such as:

CNN-LSTM Hybrid Models
Vision Transformers (ViTs)
CRNN-based OCR systems
Attention-based recognition models
Mamba and State Space Models (SSM)

Applications

Potential applications include:

Digital document preservation
Historical manuscript digitization
Assamese handwritten OCR systems
Intelligent archival systems
Educational and linguistic tools
Regional language computing

Results

The project aims to:

Improve recognition accuracy for compound characters
Build robust handwritten datasets
Develop scalable OCR pipelines
Enhance recognition performance under unconstrained settings

Future Scope

Future directions include:

Multilingual Indic OCR systems
Real-time handwritten recognition
Foundation models for document intelligence
Large-scale Assamese document digitization
Mobile-based handwritten text recognition systems

Hyperspectral Change Detection and Change Captioning

Thu, 07 Aug 2025 00:00:00 +0000

About This Project

This research project focuses on developing advanced deep learning frameworks for **Hyperspectral Change Detection ** and Change Captioning using remote sensing imagery. The work aims to automatically identify significant temporal changes in hyperspectral satellite images and generate semantic textual descriptions of detected changes.

The project explores robust AI-driven techniques combining:

Deep Learning
Vision Transformers
Mamba / State Space Models (SSM)
Attention Mechanisms
Multimodal Learning
Remote Sensing Image Analysis

The primary objective is to improve accurate land-cover change analysis, disaster monitoring, environmental assessment, and intelligent geospatial understanding from hyperspectral imagery.

Research Focus

The project investigates:

Pixel-level and semantic change detection
Temporal feature extraction from hyperspectral data
Lightweight and efficient Transformer architectures
Mamba-based sequence modeling for long-range dependencies
Vision-language models for automated change caption generation
Robustness against illumination and seasonal variations

Methodology

The proposed pipeline includes:

Hyperspectral image preprocessing and normalization
Feature extraction using Transformer and Mamba architectures
Spatial-spectral attention modeling
Change map generation
Semantic caption generation using multimodal learning frameworks

Applications

Potential applications include:

Environmental monitoring
Forest and vegetation analysis
Urban expansion detection
Agricultural monitoring
Disaster assessment
Defense and surveillance systems

Results

The project aims to achieve:

Improved change detection accuracy
Better semantic understanding of scene changes
Lightweight and scalable architectures
Enhanced interpretability in remote sensing AI systems

Future Scope

Future directions include:

Foundation models for remote sensing
Large-scale multimodal geospatial learning
Real-time change captioning systems
Edge-AI deployment for satellite analytics
Self-supervised hyperspectral representation learning

Welcoming New Lab Member Manash Choudhuri to the Research Group

Sat, 02 Aug 2025 00:00:00 +0000

We are delighted to welcome Manash Choudhuri as a new member of our research laboratory at the Department of Computer Science, Gauhati University.

Manash has joined the lab as a PhD Scholar and Junior Research Fellow (JRF) under the ISRO-RAC-S sponsored project focusing on Change Detection and Change Captioning using advanced artificial intelligence and computer vision techniques.

He brings with him two years of teaching experience along with strong research interests in Speech Processing, Deep Learning, and intelligent data-driven systems. His expertise and academic background are expected to contribute significantly to the ongoing research activities of the lab.

The entire research group warmly welcomes him and wishes him great success in his research journey and future endeavors.

Research Project Awarded on Real-World Medicinal Plant Identification Using Mobile AI Application

Thu, 20 Mar 2025 00:00:00 +0000

We are pleased to announce that a new research project has been awarded under the IERP Programme of the G.B. Pant National Institute of Himalayan Environment (NIHE) focusing on Real-World Medicinal Plant Identification using a Robust Mobile Application.

The project aims to develop an intelligent mobile-based system for accurate identification of medicinal plants in real-world environments using advanced techniques in:

Computer Vision
Deep Learning
Mobile AI Applications
Image-Based Plant Recognition
Biodiversity Informatics

The project seeks to support medicinal plant conservation, traditional knowledge documentation, and accessible AI-driven plant identification systems for researchers, students, and local communities.

The project is led by:

Dr. Sanjib Kr. Kalita, Principal Investigator (PI), Department of Computer Science, Gauhati University

with

Prof. Partha Pratim Baruah, Co-Principal Investigator (Co-PI), Department of Botany, Gauhati University

This interdisciplinary collaboration between Computer Science and Botany is expected to contribute significantly toward intelligent biodiversity management and technology-assisted medicinal plant research.

The laboratory congratulates the project team and wishes them success in this important research initiative.

A benchmark image dataset of unconstrained isolated Assamese handwritten compound characters