Research

Journals

Publication statuses and journal metrics are reported as listed in my resume.

Leveraging Multi-Agent Deep Reinforcement Learning for Autonomous Galaxy Image Classification

The Egyptian Journal of Remote Sensing and Space Sciences · SCI-E · IF: 4.1

Mentor: Dr. Don S

This study explores how multiple reinforcement learning agents can classify galaxy images. It combines agents through weighted majority voting and compares the ensemble with individual agents and deep learning baselines on the Galaxy Zoo dataset.

  • Developed DRL agents using DQN, PPO, Recurrent PPO and A2C with customized policies and feature extractors for galaxy image classification.
  • Combined agents through weighted majority voting for efficient classification.

Enhancing Yoga Pose Recognition Through a Multi-Modal Hybrid Approach with Attention Mechanism and Late Fusion on YFA-30

International Journal of Yoga · SCI-E · IF: 1.1

Mentor: Dr. Don S

This work combines body landmarks and visual features to recognize yoga poses. The hybrid approach brings together MoveNet, ResNet50, attention and late fusion, and introduces YFA-30 as a dataset for evaluating yoga pose recognition.

  • Combined MoveNet’s efficient pose detection with ResNet’s feature extraction, attention mechanisms and late fusion for robust yoga pose classification.
  • Introduced YFA-30, a benchmark dataset designed for yoga pose recognition.

Adaptive Gaze Estimation Through Modified Deep Learning and XGBoost Fusion

SN Computer Science · Under review

Mentor: Dr. Don S

This work investigates a hybrid approach to estimating eye gaze using a modified VGG16 network and an XGBoost classifier. It evaluates the approach across variations in lighting and eye appearance, extending earlier research on VGG16–XGBoost gaze estimation.

  • Developed a modified VGG16–XGBoost gaze estimation system for datasets with varied lighting, eye appearance and dimensionality.
  • Outperformed the compared deep learning and machine learning models, demonstrating potential for real-time gaze estimation.
  • Extended the work titled “Gaze Estimation Using VGG16 Architecture Through XGBoost Classifier”.

Harnessing Deep Reinforcement Learning Algorithms for Image Categorization: A Multi Algorithm Approach

Engineering Applications of Artificial Intelligence · SCI-E · IF: 8.0

Mentor: Dr. Don S

This study evaluates reinforcement learning as an approach to image categorization across multiple benchmark datasets. It compares different DRL algorithms with conventional machine learning and deep learning methods to assess their classification performance.

  • Investigated multiple DRL algorithms for image categorization.
  • Demonstrated the potential of DRL for real-time computer vision challenges.

Conferences

26 February 2024

Analysis of Image Restoration Techniques on Lunar Surface Images

2023 Innovations in Power and Advanced Computing Technologies (i-PACT)

This paper examines noise removal in lunar surface images captured by Chandrayaan-II. It compares Wiener, median and Gaussian filtering across different noise conditions, evaluating restoration quality using measures including SNR, PSNR and mean squared error.