🦖 PaleoNet: Dinosaur Species Classifier

🌟 Overview

PaleoNet is an advanced dinosaur species classification system built using deep learning. The application can identify 15 different dinosaur species from images with high accuracy, using a convolutional neural network (CNN) based on EfficientNetB0 architecture.

Explore prehistoric creatures through the power of artificial intelligence! 🔍

✨ Features

• Image Classification - Upload your own dinosaur images for instant species identification
• Interactive Gallery - Explore sample images from the test dataset
• Species Encyclopedia - Learn fascinating facts about each dinosaur species
• Model Insights - Visualize the model architecture and performance metrics
• Tabbed Navigation - Switch easily between Home, Upload Image, and Sample Gallery sections

🖼️ Application Screenshots

Main Classification Page: Upload and classify dinosaur images

Dinosaur Encyclopedia: Learn about different dinosaur species

Model Information: Visualize the model architecture and performance

🚀 Quick Start

Install requirements

pip install -r requirements.txt

Run the Streamlit app

streamlit run PaleoNet.py

Once started, the application will be available at http://localhost:8501

🦕 Supported Dinosaur Species

PaleoNet can classify the following 15 dinosaur species:

🦖 Ankylosaurus	🦕 Brachiosaurus	🦎 Compsognathus
🦖 Corythosaurus	🦖 Dilophosaurus	🦎 Dimorphodon
🦖 Gallimimus	🦎 Microceratus	🦖 Pachycephalosaurus
🦖 Parasaurolophus	🦖 Spinosaurus	🦕 Stegosaurus
🦕 Triceratops	🦖 Tyrannosaurus Rex	🦖 Velociraptor

📂 Project Structure

PaleoNet/
├── app/                    # Application code
│   ├── app.py              # Main Streamlit application
│   ├── utils.py            # Utility functions
│   └── pages/              # Additional app pages
│       ├── 01_Model_Info.py
│       └── 02_Dinosaur_Encyclopedia.py
├── assets/                 # Images and static assets
├── data/                   # Dataset directory
│   └── dinosaur_dataset_split/
│       ├── train/          # Training data (70%)
│       ├── val/            # Validation data (15%)
│       └── test/           # Test data (15%)
├── docs/                   # Documentation
├── model/                  # Saved model files
│   ├── dinosaur_classifier_transfer_learning.keras
│   ├── dinosaur_class_mapping.json
│   └── dinosaur_model_performance.json
├── notebooks/              # Jupyter notebooks
│   └── opdracht_CNN_stijnen_simon.ipynb  # Model training notebook
├── requirements.txt        # Project dependencies
├── README.md               # Main documentation
└── LICENSE                 # MIT License

📊 Model Architecture

PaleoNet uses a transfer learning approach based on EfficientNetB0:

• Base Model: EfficientNetB0 pre-trained on ImageNet
• Feature Extraction: Global Average Pooling to reduce spatial dimensions
• Classification Head:
  • Dense layers (512 & 256 neurons) with ReLU activation
  • Batch Normalization for training stability
  • Dropout (0.4) for regularization
  • Output layer with 15 neurons and softmax activation
• Training Enhancement: Data augmentation to improve generalization

The model achieves over 70% accuracy on the test set, with especially strong performance on distinctive species.

📝 Documentation

Visit the docs directory for detailed documentation:

• User Guide
• Model Information
• Development Guide

💻 Streamlit Application

PaleoNet uses Streamlit to create an interactive web application for dinosaur image classification. The application includes:

• Main Page: Upload your own images or try sample images for classification
• Model Info Page: Visualize the model architecture and performance metrics
• Dinosaur Encyclopedia: Learn fascinating facts about each dinosaur species

Key Application Features

• Real-time classification with confidence scores
• Top-3 prediction display
• Interactive sample gallery with random test images
• Performance metrics visualization
• Detailed dinosaur information cards

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgements

• Dataset: Dinosaur Image Dataset from Kaggle
• Technologies:
  • TensorFlow and Keras for deep learning
  • EfficientNetB0 architecture by Google
  • Streamlit for the web application interface
  • Matplotlib and Seaborn for visualizations
• Paleontological Information: Various academic sources for dinosaur facts and information
• Development: Simon Stijnen for the AI Deep Learning course at VIVES University of Applied Sciences (2025)