Projects
A selection of my research and personal projects. More on GitHub.
Jetson IMX708 CUDA/WebRTC Camera Pipeline
A headless camera pipeline for the Sony IMX708 on NVIDIA Jetson Orin Nano. It captures raw RG10 Bayer frames with V4L2, processes them through a persistent CUDA ISP, and streams low-latency 720p H.264 video over WebRTC. The pipeline includes automatic exposure, white balance, color correction, autofocus, and server-side MP4 recording.
Python CUDA Jetson Computer Vision WebRTC
Computer Vision Codes
Collection of computer vision implementations and experiments including image processing, object detection, and segmentation pipelines.
Jupyter Notebook Python Computer Vision
FPV Simulation
FPV drone simulation with physics, input handling, and rendering built in TypeScript/Vite.
TypeScript Vite Physics Simulation
React ROS Control Panel
A web-based control panel built with React for controlling ROS (Robot Operating System) robots. Provides real-time robot telemetry and control interface.
JavaScript React ROS
GoPro ROS
ROS integration package for GoPro cameras, enabling image acquisition from GoPro devices for robotics and data collection applications.
Python ROS Robotics
DWA Simulator
Python implementation of the Dynamic Window Approach (DWA) for robot path planning and local navigation.
Python Path Planning Robotics
River Pollution Tracker
IoT-based river water pollution monitoring system using ESP32 microcontrollers for real-time environmental data collection.
HTML ESP32 IoT
Precision Egg Counting System
Deployed a deep learning pipeline to count eggs on a Raspberry Pi 5 without GPU, achieving 5 FPS inference speed using optimized models.
Python PyTorch Raspberry Pi
Grape Segmentation in Vineyard using UNet Architecture
Thesis Project. Implemented an end to end pipeline to train a UNet model for grape segmentation from images collected by a Husky unmanned ground vehicle.
Python UNet Image Segmentation Robotics
Pod Detection in Edamame
Implemented an advanced machine learning course project on edamame pod detection and segmentation, leveraging the state-of-the-art Mask R-CNN model to achieve precise and efficient object localization and instance segmentation.
Python TensorFlow Keras Mask R-CNN YOLOv5
Tomato Leaf Disease Classification
Comparison of different classification models for identifying unhealthy and healthy tomato leaves. Trained and tested AlexNet, GoogleNet, and MobileNet models on the PlantVillage dataset.
Python Google Colab Image Classification
Covid 19 Information App
Covid-19 information mobile application built during the lockdown period, providing pandemic-related information in the Nepali language.
Flutter Mobile App Development