RockHound 🔍🪨

RockHound is a pocket geologist app designed for outdoor enthusiasts, budding geologists, and anyone with an interest in rocks and minerals. Users can snap a picture of any rock or mineral they find, and the app will use a custom-trained AI model to identify it, providing key information and details about its characteristics.

Features

• AI-Powered Rock Identification: RockHound features a custom AI model designed to accurately identify rocks and minerals from images. By leveraging machine learning, the app becomes a powerful tool for identifying and understanding the world of geology.

• Interactive AI Chatbot (Rocco the RockHound): Engage with Rocco the RockHound, the interactive AI chatbot. Rocco answers your geology-related questions and makes learning about rocks fun!

• Digital Collection: Maintain a digital collection of your finds. Users can save identified rocks, revisit them later, and learn more about their geological features.

Project Focus

The project has evolved to include not only the development of the RockHound mobile app but also the training and refinement of a machine learning model specifically for rock identification. This model is trained using a diverse dataset of rock images, employing techniques such as data augmentation and convolutional neural networks (CNNs) to achieve high accuracy in identifying various rock types.

Setup (To run model)

Disclaimer:

Training a convolutional neural network can be resource intensive!

1. Environment

You'll want to have the following installed (Skip if using Google Colab):

• Python version 3.8 - 3.11

The Rockhound convolutional neural network(CNN) was written in a Jupyter Notebook. To work with the .ipynb file:

• Install a Jupyter extension for your IDE of choice (additional Jupyter related extensions may be required). If you're using JupyterLab or Google Colab this can be skipped.

2. Download Project Files

Once your environment is setup, download the ZIP containing all the files. For this purpose you'll only need the files within the rh_model folder. This should contain:

• dataset, a folder containing all the data required to train/test the model.
• rockhound_cnn.ipynb, a file containing the model code.

3. Setup Dependencies

Open rockhound_cnn.ipynb and prior to running the first cell select a python kernel (must be between versions 8 - 11). Then, running the first code cell installs all dependencies necessary (this may take a minute). After this has completed, all imports needed are listed throughout the notebook, simply run each cell.

4. Run Notebook Cells

Lastly, run the reamaining cells in sequential order observing the model's training, evaluation, and results.