🌌 Cosmic Navigators – AI-Powered Space Debris Detection System A capstone project aimed at building a lightweight, end-to-end system for detecting, classifying, and tracking space debris from user-uploaded images and videos using cutting-edge AI and cloud technologies.

🚀 Project Overview

Problem:

    Space debris poses a growing risk to satellites and spacecraft, yet monitoring tools are often inaccessible to non-experts.

Our Solution:

    A full-stack, web-based platform that leverages computer vision to detect and classify space debris in real-time.

🛠️ Tech Stack

    Layer	Tools/Technologies
    Frontend	React.js (hosted on AWS S3)
    Backend	FastAPI (Dockerized, hosted on AWS EC2)
    AI Models	YOLOv8 (detection), ResNet50 (classification)
    Deployment	Docker, AWS EC2, AWS ECR, AWS S3

🎯 Features

    Upload images or videos to detect space debris

    Real-time bounding box predictions and classification

    Adjustable confidence threshold for results

    Export results as annotated images or PDF reports

    Stream video with live annotation

    Bonus: Space-themed chatbot for space data

🧠 Model Architecture

    YOLOv8: For real-time object detection (bounding boxes)

    ResNet50: Transfer learning for classifying debris types

    Custom Tracker: Logic-based tracking using bounding box coordinates

🔢 Evaluation Metrics

    Model	Metric	Result
    YOLOv8	mAP	~88%
    ResNet50	Accuracy	~90%
    Inference Time	Avg/image	~0.8 seconds

📦 Setup Instructions

Clone the Repository

    bash
    Copy
    Edit
    git clone [REPO_URL]
    cd cosmic-navigators
    Run Docker Backend

    bash
    Copy
    Edit
    docker build -t debris-api .
    docker run -p 8000:8000 debris-api
    Run Frontend

    Deploy React app to S3 static hosting

    Configure API endpoints in .env

📂 Repository Contents

    backend/: FastAPI app, YOLO & ResNet models

    frontend/: React app interface

    models/: Trained model weights

    deployment/: Dockerfiles, AWS setup

    notebooks/: Training and evaluation scripts

🌍 Reproducibility

    Clone the repo and follow setup steps

    Models load automatically from /models

    Supports API requests via Postman or UI