• Piloted a drone equipped with thermal imaging cameras to capture high-resolution thermal images of buildings, covering over 50 buildings within a single month to compile a diverse dataset for analysis.

• Leveraged the YOLOv4 architecture to develop a deep learning model capable of identifying heat envelopes around windows with precision, reducing analysis time from 3-5 days to under 30 minutes per building.

•  Analysed thermal data to provide actionable insights for building insulation improvements, contributing to an average of 20% reduction in energy waste across evaluated buildings.

• The project's success and methodologies were detailed in a paper for IEEE AIPR 2023, setting a new standard for rapid and accurate energy efficiency auditing using Deep learning techniques.

•Developed a comprehensive system integrating CO2, temperature, and humidity sensors with a YOLOv4 deep learning model to analyse classroom occupancy in real-time, leveraging Python for seamless data synchronization.

•Created and fine-tuned a YOLOv4 model to detect occupancy with a confidence level of 98%, using over 20,000 images from classroom settings to ensure accuracy and reliability in diverse educational environments.

•Engineered a smart HVAC control algorithm that adjusts based on real-time occupancy and environmental data, achieving up to 30% improvement in energy efficiency and enhancing thermal comfort for occupants.

•Presented the project at IEEE South East Conference 2024, demonstrating the potential of AI in smart building management and contributing to sustainability goals by optimizing energy consumption.

Emg sensors placed on different muscle groups

• Designed and executed an innovative experiment by placing EMG sensors across multiple muscle groups on the hand, capturing dynamic muscle activity data during various gestures for comprehensive analysis.

•Utilized Python to develop two distinct machine learning models, Random Forest and Logistic Regression, training on a dataset comprising over 8 hand gesture instances to fine-tune model parameters for optimal performance.

•Achieved a landmark accuracy of 98% in gesture recognition, outperforming existing benchmarks by 12% through meticulous data preprocessing and feature engineering, ensuring robustness across diverse environmental conditions.

•Authored a paper on the findings and methodology, contributing to the IEEE AIPR 2023 conference, highlighting the potential of EMG data in developing advanced HCI systems.

•Developed a real-time face mask detection system using the MobileNet-V2 CNN architecture, processing live video feeds to detect mask presence with a 96% accuracy rate, crucial during the COVID-19 pandemic. 

•Compiled a comprehensive dataset of over 15,000 labeled images featuring individuals with and without masks under various lighting conditions and angles to train the deep learning model.

•Deployed the model in public spaces to encourage mask-wearing, contributing to community health safety efforts and demonstrating the practical application of AI in public health emergencies.

•Awarded the Best Poster at a graduate college competition, acknowledging the project's innovative approach and impact on promoting safety measures during the pandemic.