Embedded Files
Digital Skills
Digital Skills
Programming: Python, C, C++, Java, JavaScript, PHP, Bash, Assembly, Dart, HTML, CSS, Node.js
Virtualization: VMware, VirtualBox, AWS, Firebase
Libraries: NumPy, Pandas, Matplotlib, Seaborn, SciPy, Scikit-Learn, React
Visualization: PowerBI, Tableau, Microsoft PowerPoint
Writing: LaTeX, Microsoft Word, Microsoft Excel, Notion
Frameworks: TensorFlow, PyTorch, PyCharm, OpenGL, Flask, Django, Flutter, Laravel, Bootstrap, Unity
Databases: Oracle, Microsoft SQL Server, MySQL, MongoDB
Simulation: MATLAB, PSpice, Packet Tracer, LTSPICE, Proteus, Cisco Packet Tracer, JavaFX
OS: Linux (Ubuntu, CentOS), Windows
Hardware: Raspberry Pi, Arduino
Languages: English, Bengali, Hindi
My Current Projects
Multitasking learning approach for Wildfire Classification and Segmentation
Wildfires are vital for ecosystems but threaten human lives and the environment, particularly in remote areas. Early detection is challenging due to environmental factors. This study uses satellite data and UAVs to leverage machine learning with the Involution Neural Network (Inv-Net) to enhance wildfire segmentation and classification. Inv-Net improves feature extraction and spatial correlation, integrating classification and segmentation for better accuracy. The Dark Channel Prior (DCP) technique mitigates atmospheric disturbances, enhancing image clarity. Our approach demonstrates robust wildfire detection and mapping in difficult conditions.
Deep Learning Architectures for Early Detection of Lung Diseases in Chest X-ray Images
Our study introduces a deep learning system for detecting COVID-19 and viral pneumonia from chest X-rays. Using the COVID-QU-Ex dataset, we developed an ensemble model combining five pre-trained models and integrated explainable AI methods for transparency. We also created a Flask web application for easy access and immediate analysis of X-ray images, enhancing early lung disease detection and revolutionizing diagnostics.
Review on the Usage of Machine Learning Models and their Futures
Machine learning has revolutionized fields such as healthcare, e-commerce, and education, enabling advancements like self-driving cars, sophisticated diagnostics, real-time language translation, and personalized recommendations. This study explores the profound impact of machine learning, highlighting its capabilities and the ethical challenges it presents, including data privacy, algorithmic bias, and transparency. The paper includes a detailed methodology, a comprehensive review of selected studies, key findings, and a discussion on future opportunities and limitations for ethical and responsible ML use.
Academic Projects
Academic Projects
1. Optimizing Patient Feedback with Generative Adversarial Network Leveraging Knowledge Distillation to Improve Healthcare (Ongoing)
Description:
Developing a patient feedback system using a transformer-based GAN with knowledge distillation to efficiently analyze reviews of Dhaka’s private hospitals. This system aims to enhance local healthcare quality and transparency by providing actionable insights using advanced Generator and Discriminator components within the model framework.
Tech Stacks:
Deep Learning Frameworks: TensorFlow, PyTorch
Models: Generative Adversarial Networks (GAN), Transformer-based Models
Other Tools: Knowledge Distillation
2. Enhancing Intrusion Detection: A Robust Ensemble Learning Approach with Explainable AI
Description:
Developed an ensemble model for network intrusion detection on the UNSW-NB15 dataset, combining Decision Tree, Random Forest, Naive Bayes, Perceptron, and Logistic Regression. The model uses majority voting and stacking with XGBoost as a meta-classifier and integrates LIME for interpretability.
Tech Stacks:
Machine Learning Libraries: Scikit-learn, XGBoost
Techniques: Ensemble Learning, Explainable AI (LIME)
Datasets: UNSW-NB15
3. Federated Learning for Secure Medical Image Analysis with BLIP Model Integration
Description:
Integrated the BLIP model within federated learning frameworks for vision-language tasks in medical imaging. Applied homomorphic encryption to secure CT scans and chest X-rays, enabling encrypted image analysis and question-answering while preserving privacy.
Tech Stacks:
Deep Learning Frameworks: PyTorch, TensorFlow
Models: BLIP Model, Federated Learning
Encryption Techniques: Homomorphic Encryption
Data Types: Medical Imaging (CT scans, Chest X-rays)
4. Buffer Overflow Exploitation and Mitigation with CSRF Attack Simulation
Description:
Exploited a buffer overflow vulnerability using SEEDUbuntu12 to gain root privileges, followed by implementing OS-level protections against such attacks. Conducted a CSRF attack on a social networking app, showing how a malicious site could inject unauthorized requests into a trusted session.
Tech Stacks:
Security Tools: SEEDUbuntu, OS-level Protection Techniques
Vulnerabilities: Buffer Overflow, CSRF Attacks
Programming Languages: C, Python, Bash
5. UAV Cyber-Physical Intrusion Detection System with Federated Learning
Description:
Developed an IDS for UAVs by simulating cyber-attacks and collecting cyber-physical data. Optimized the model with Whale Optimization and Knowledge Distillation for efficient detection in Federated Learning, using different neural networks as local classifiers.
Tech Stacks:
Machine Learning: Federated Learning, Knowledge Distillation
Optimization Algorithms: Whale Optimization
Data Sources: Cyber-Physical Data (UAV simulations)
Frameworks: TensorFlow, PyTorch
6. Autonomous Vehicle Obstacle Detection by Observing Potholes
Description:
Developed autonomous vehicle navigation in South Asian road conditions using YOLO-based models for real-time obstacle detection. Implemented the solution with Python, TensorFlow, and Arduino, providing robust performance in diverse environments.
Tech Stacks:
Computer Vision: YOLO (You Only Look Once)
Frameworks: TensorFlow
Embedded Systems: Arduino
Programming Languages: Python, C++
Page updated
Google Sites
Report abuse