Computer Vision .
Calorie Counter App
An intelligent calorie counter app that accurately detects food items and extracts weight data from LCD screens, providing users with precise calorie calculations.
Real time 2D Object Recognition
An object detection system utilizes classical computer vision techniques, without relying on deep learning networks.
Customized Region Proposal ...
Road Surface Segmentation
A novel training method for road surface segmentation that improves adaptability across diverse environments for autonomous driving
Developed a traffic sign detection system that integrates a customized object proposal network, to improve the frame rate.
Content based image Retrieval
An image matching system that uses feature-based comparisons like color, texture, and deep networks, along with various distance metrics for accurate similarity detection.
AI & ML .
A comparative analysis of traditional Perturb & Observe and ANN-based MPPT for solar PV systems, showcasing improved tracking efficiency and stability under varying conditions with intelligent MPPT.
Comparison of Conventional and Intelligent MPPT Implementation using Boost Converter
Brain Computer Interface
A BCI system that accurately detects body part stimulations from EEG signals using advanced deep learning techniques.
Analysis of Deep Networks for Recognition
Explored how to manipulate and interpret deep networks. Worked with training, validation, and custom test datasets of our own
Robotics .
Mapping Streets of Boston
LiDAR SLAM-based mapping of Boston streets using an autonomous car, with LeGO-LOAM and IMU integration for precise pose estimation.
Reconnaissance Bot
A ROS package for bots exploring unknown environments using frontier exploration and adaptive path planning on the TurtleBot3 platform.
Sensor Fusion for Autonomous Vehicle
IMU and GPS data is used to perform sensor fusion, for improve accuracy in tracking position, orientation, and velocity
Neural Network based Path Planning
Local Path Planning of a differential drive robot in a race track, using deep neural networks and CNNs.