PhD Student in Computer Science | Researcher |
Description: Developed a railway fault detection system using transfer learning on a large image dataset to identify faults in railway tracks. The model achieved high accuracy, significantly reducing the manual inspection time.
Technologies: TensorFlow, Keras, Python, Transfer Learning (ResNet50)
Challenges: Handling imbalanced data, fine-tuning the model for optimal accuracy, and ensuring real-time deployment on edge devices.
Outcome: The model can predict faults with over 90% accuracy, helping to automate fault detection in real-time.
Description: Implemented a pothole detection system using deep learning models on 2D LiDAR data. This system allows real-time detection of potholes on roads, enabling municipalities to prioritize maintenance effectively.
Technologies: Python, OpenCV, Keras, LiDAR (2D), Deep Learning (CNN)
Challenges: Preprocessing LiDAR data for feature extraction and improving the model’s ability to generalize across different environments.
Outcome: The solution was deployed in a pilot project for smart city development, providing real-time pothole detection with high precision.
Description: Built a face detection system using deep learning models on the Jetson Nano for edge computing. The system can detect and track faces in real-time, making it suitable for surveillance and security applications.
Technologies: Jetson Nano, OpenCV, Python, Deep Learning (Haar Cascades, CNN)
Challenges: Optimizing the system for real-time performance on edge devices and ensuring accurate face tracking in varied lighting conditions.
Outcome: The system provides real-time face detection and is deployed in a local security setup, with plans for further integration with smart cameras.
Description: Developed a personalized shopping website using AWS services like Amazon Personalize and EC2 for scalable recommendations. The website tailors product suggestions based on user behavior and preferences.
Technologies: AWS (Amazon Personalize, EC2), JavaScript, Python, Flask
Challenges: Ensuring scalability, optimizing the recommendation system, and integrating it seamlessly into the website's UI.
Outcome: The platform is now live with thousands of active users, providing personalized shopping experiences that drive customer engagement and sales.
Description: Designed and developed a smart dustbin system equipped with various sensors and an autonomous vacuum cleaning mechanism. The dustbin has an ultrasonic sensor at the lid to automatically open it when someone approaches, allowing easy disposal of waste. It also features a vacuum cleaner at the bottom, which can be remotely controlled to clean dust and debris around the bin.
Technologies: IoT, Raspberry Pi, Ultrasonic Sensors, Gas Sensors, Fire Sensors, Buzzer, Python
Key Features:
Challenges: Integrating various sensors and ensuring seamless communication between components, optimizing the system for power efficiency, and ensuring the vacuum mechanism was reliable in a real-world environment.
Outcome: The system was successfully developed and demonstrated as an innovative solution for smart waste management. It is designed to enhance sanitation and safety in urban areas, with potential applications in smart cities and automated cleaning systems.
Description: Developed a smart wheelchair that can navigate autonomously, avoiding obstacles and detecting environmental changes. It also provides real-time data about the wheelchair’s status to caregivers through IoT integration.
Technologies: IoT, Raspberry Pi, ROS, Machine Learning
Challenges: Ensuring the wheelchair’s autonomy in complex environments and ensuring low-latency communication between devices.
Outcome: The system is currently being trialed in healthcare facilities, assisting individuals with limited mobility and improving their independence.