Transfer Learning for Geological Images

Description

This is the repository for the code in Dawson et al 2022. This code can be used to easily to perform transfer learning for geological images using any of the built-in TensforFlow-Keras image classification models used in the paper.

Built-In Models

The models available here are:

Installation

This project has the following dependencies:

• Numpy sudo pip install numpy
• OpenCV Python sudo apt-get install python-opencv
• TensorFlow sudo pip install --upgrade tensorflow-gpu
• Keras sudo pip install tf.keras

Usage

To use this code, please ensure you have set up your dataset in folders following this structure:

├── "dataset_name"                   
|   ├── train
|   |   ├── class_1_images
|   |   ├── class_2_images
|   |   ├── class_X_images
|   |   ├── .....
|   ├── val
|   |   ├── class_1_images
|   |   ├── class_2_images
|   |   ├── class_X_images
|   |   ├── .....
|   ├── test
|   |   ├── class_1_images
|   |   ├── class_2_images
|   |   ├── class_X_images
|   |   ├── .....

Then you can simply run combined-model.py.