CoolFace
Apppublic

Mohiit007/BrainCache-OrbitalScan

sourceHugging Faceupdated 1y agoView on Hugging Face
0likes
App README

BrainCache - OrbitalScan

AI-powered detection of Toolbox, Oxygen Tank, and Fire Extinguisher to ensure astronaut safety.

Features

  • —Upload Image, Video, or Use Live Camera
  • —YOLOv8-based object detection
  • —Confusion Matrix and mAP analytics
  • —PDF Performance Report Generation
  • —Space-themed Streamlit UI

How to Run Locally

bash
pip install -r requirements.txt
streamlit run app.py


# 🚀 OrbitalScan: Space Station Object Detection (YOLOv8)

**OrbitalScan** is an AI-powered computer vision solution designed to **detect critical objects inside a space station** using **YOLOv8**.  
Developed for **BuildWithDelhi 2.0 Hackathon** by **Team BrainCache**.

---

## 1. Problem Statement
Astronauts need **real-time monitoring of essential tools** (Toolbox, Oxygen Tanks, Fire Extinguishers) to ensure safety and operational efficiency in space missions.

---

## 2. Our Solution
We trained a custom **YOLOv8 model** achieving:
- **mAP@0.5 = 0.916**
- **mAP@0.5-0.95 = 0.792**

The solution is deployed as a **Streamlit Web App** with:
- Image/Video Upload
- Live Camera Detection
- Annotated Results Download
- Confusion Matrix & Performance Metrics
- Auto-generated PDF Reports

---

## 3. Features
- **High Accuracy** (91.6% mAP@0.5)
- **Interactive Space-Themed UI**
- **Live Detection via Camera**
- **PDF Performance Reports**
- **Optimized YOLOv8 Model**

---

## 4. Technologies Used
- **Python** (YOLOv8, OpenCV, Streamlit)
- **PyTorch** for model training
- **Pandas & Matplotlib** for analysis
- **FPDF** for report generation
- **Streamlit-Lottie** for animations

---

## 5. Project Structure
OrbitalScan/
│── app.py # Streamlit Web App
│── train.py # Training Script
│── predict.py # Inference Script
│── best.pt # Trained YOLOv8 Model
│── data.yaml # Dataset Configuration
│── requirements.txt # Dependencies
│── results/ # Performance Results (results.png, confusion matrix, etc.)
│── README.md # Documentation

yaml
Copy
Edit

---

## 6. How to Run Locally
1. **Clone the Repository**

git clone https://github.com/YourRepo/OrbitalScan.git cd OrbitalScan Install Dependencies

bash Copy Edit pip install -r requirements.txt Run the App

bash Copy Edit streamlit run app.py

  1. 1.Deploy on Streamlit Cloud Push this repo to GitHub

Go to Streamlit Cloud

Connect your repo and click Deploy

App will be available at:

arduino Copy Edit https://your-app-name.streamlit.app

  1. 1.Performance Report mAP@0.5: 0.916

mAP@0.5-0.95: 0.792

Confusion Matrix: Included in results/

Failure Case Analysis: Low-light images caused minor misclassifications. Plan: Add more data augmentation.

  1. 1.Demo Web App: Streamlit App URL

Presentation: Google Slides

Model Weights: Google Drive

  1. 1.Team BrainCache

Swastika

Mohit

Uday

Rohit

  1. 1.License MIT License.
  1. 1.Acknowledgements Special thanks to BuildWithDelhi 2.0 Hackathon organizers and Ultralytics YOLO.