Research Papers and Publications
1
Experiments in Single Image Super Resolution using Convolutional Neural Networks and Auto-encoders
The Super-Resolution Convolutional Neural Network (SRCNN) is utilized in PyTorch to enhance image quality, achieving a testing PSNR of 26.117 and training SSIM of 91.6% through the perceptual Mean Gradient Error (MGrE) loss function. An auto-encoder is also implemented, obtaining a max PSNR of 33 on the DIV2K dataset, demonstrating improved image reconstruction.
2
Face recognition and detection using CNN
Face recognition and detection using CNNs involve training a model to identify faces in images by extracting features through convolutional layers. This technique enables real-time detection and classification, with applications in security, access control, and social media tagging.
3
Effect of dry carbon fibre on the mechanical properties of G801 carbon fibre reinforced composite
The field of fiber-reinforced composite materials advances rapidly, thanks to their lightweight and superior properties compared to conventional materials. This study focuses on fabricating carbon fiber reinforced composites using the lay-up procedure in various sequences, optimizing production costs by sandwiching dry fibers with pre-pregs while ensuring no defects or performance compromise.
4
Design and Development of a Multi-terrain Vehicle
This report discusses the Rocker-Bogie mechanism used in Mars exploration rovers, comparing six-wheel and eight-wheel designs. It evaluates the rover's ability to navigate obstacles over twice the wheel diameter under Martian conditions for potential emergency supply and route determination applications.
Academic Projects
01
Real-Time Emotion Detection using CNN (PYTHON)
The Real-Time Emotion Detection project employs Convolutional Neural Networks (CNN) to analyze facial expressions from video streams. By classifying emotions such as happiness, sadness, and anger, it enables applications in user interaction, security, and mental health monitoring through instantaneous feedback.
03
Computer Vision: Real time Angry bird trajectory tracking (PYTHON)
Real-time Angry Bird trajectory tracking with computer vision involves capturing frames, detecting the bird using object detection methods like edge detection or color thresholding, and applying feature tracking (optical flow). Kalman filters or polynomial regression predict its path based on real-time velocity and position data. This enables accurate flight trajectory prediction for interactive gaming and performance analysis applications.
05
Ultrasonic Braking System
An ultrasonic braking system uses ultrasonic sensors to detect obstacles and send signals to the vehicle's control unit, triggering the brakes automatically. This system enhances safety by providing real-time obstacle detection, reducing reaction time, and preventing collisions.
07
Modelling of Grid Frequency Stabilization by Utilizing Synthetic Inertia in Wind Turbine (MATLAB)
The modeling of grid frequency stabilization using synthetic inertia in wind turbines involves simulating how wind turbines can emulate the inertial response of conventional power plants. By adjusting power output during grid frequency deviations, synthetic inertia helps mitigate frequency fluctuations, enhancing grid stability in renewable energy-integrated systems.
02
Path Tracking, Line Following, Obstical Detecting robot using SLAM (PYTHON)
The Path Tracking, Line Following, and Obstacle Detecting robot employs SLAM (Simultaneous Localization and Mapping) on a Raspberry Pi, and ultrasonic sensors for environment mapping and obstacle detection. Using algorithms like A* for path planning and PID controllers for line following, the robot processes real-time sensor data to navigate autonomously, adapt to dynamic environments, and maintain accurate positioning.
04
Face recognition using Eigenface (MATLAB)
Face recognition using the Eigenface method involves decomposing facial images into principal components through Principal Component Analysis (PCA). These components, called Eigenfaces, capture key facial features, enabling efficient comparison and recognition by projecting new faces onto this reduced feature space.
06
Design of a Die for Plastic Moulds on Auto-CAD
Designing a die for plastic molds in AutoCAD involves creating detailed 2D and 3D models of the core, cavity, parting line, gating, ejection system, cooling channels, and draft angles, ensuring precise dimensions for efficient molding and part removal.
08
FPGA based display system with keyboard interfacing on Digilent Nexys4-DDR
An FPGA-based display system on the Digilent Nexys4-DDR interfaces with a keyboard using PS/2 protocol, processing input signals and driving the VGA display in real-time for user-controlled visual output.