PaleoNet
PaleoNet application interface showing dinosaur image classification results
Demo of PaleoNet classifying a dinosaur image with CNN model predictions
Training progress visualization showing model accuracy and loss curves over epochs
Model performance metrics including accuracy, precision, recall, and confusion matrix
Project Overview
A machine learning assignment that turned into an engaging deep learning challenge. Working with a dataset of dinosaur images, I built PaleoNet to automatically classify different dinosaur species using convolutional neural networks, with the goal of achieving above 70% accuracy.
The most rewarding part was optimizing the model architecture and seeing the accuracy improve to just above 80% through careful data preprocessing and hyperparameter tuning. I experimented with different CNN architectures, implemented data augmentation techniques to handle the limited dataset, and deployed the final model as an interactive Streamlit application.
This project taught me the importance of systematic experimentation in AI, methodically testing different approaches and analyzing the results to understand what works and why.
Technologies Used
- AI
- CNN
- TensorFlow
- Streamlit
- Keras
- Python
- NumPy
- Pandas