Space Debris Detection 🚀 Overview 🌌 This project trains a YOLOv8 model to detect space debris and satellites in images! 📸 It uses the Roboflow platform to download a dataset and Ultralytics YOLO for object detection. Perfect for space enthusiasts! 🪐 Features ✨
Detects Space Objects: Identifies debris and satellites like CHEOPS, SOHO, and more 🛰️ YOLOv8 Model: Trains a lightweight YOLOv8n model for fast detection ⚡ Roboflow Integration: Downloads the "Space-debris-2" dataset automatically 📥 Training Metrics: Tracks box loss, classification loss, and more 📊 Image Prediction: Tests the model on sample images with confidence scores 🖼️
Requirements 🛠️
Python 3.12+ 🐍 Libraries: roboflow 🌐 ultralytics (YOLO) 🔍 opencv-python-headless 📷 numpy 🔢 matplotlib 📈 Pillow 🖼️
Installation ⚙️
Clone the repo:git clone https://github.com/your-username/space-debris-detection.git cd space-debris-detection
Install packages:pip install roboflow ultralytics opencv-python-headless numpy matplotlib Pillow
Get a Roboflow API key from Roboflow 🔑
Usage 🎯
Update the Roboflow API key in the notebook:rf = Roboflow(api_key="your-api-key-here")
Run the Jupyter notebook:jupyter notebook space_debris_detection.ipynb
The notebook will: Download the "Space-debris-2" dataset 📦 Train a YOLOv8n model for 10 epochs 🧠 Save results in runs/detect/trainX 📁 Predict on a sample image and show results 🖥️
View training plots and predictions in the notebook or runs/detect folder 📈
File Structure 📁
space_debris_detection.ipynb: Main notebook for dataset download, training, and prediction 🖥️ README.md: This guide! 📝 Space-debris-2/: Downloaded dataset folder (auto-generated) 📷
Notes 📌
Training runs on CPU; GPU can speed it up if available ⚡ Dataset includes 2721 training and 239 validation images 🖼️ Model detects 11 classes (e.g., debris, CHEOPS, SOHO) 🌠 Prediction confidence threshold is set to 0.25 by default 🎯 Check data.yaml in the dataset folder for configuration 🔍
Troubleshooting 🐛
API key error?: Verify your Roboflow API key 🔑 Dataset not found?: Ensure internet connection for download 🌐 Slow training?: Reduce batch size or use a GPU 🖥️ Prediction issues?: Check image path and model weights in runs/detect 📸
License 📜 MIT License – free to use and share! 😊