hutton

This project aims to classify rock types using image classification methods powered by Tensorflow.

Infrastructure

• Python 3.7

Required Modules for Standalone Desktop Utilization

python pip install os
python pip install tensorflow
python pip install numpy
python pip install keras

Fundamental Use of Hutton Pipeline

• You can get the dataset from Kaggle. Feel free to contribute :D

Prepare a Image Dataset for Hutton

• [Required] Prepare train and validation datasets to make Hutton model

# Initialize the dataset instance
hutton_v1_dataset = Hutton_Dataset()
# You can use your own data by just parameterizing the absolute directory
data_dir = "D:/GitRepos/hutton/rock_samples/train"
hutton_v1_dataset.set_DATA_DIR(data_dir)
# You can also use external data sources that are zipped
external_data_url = 'https://github.com/Quelich/hutton/blob/main/rock_samples/train.zip'
hutton_v1_dataset.set_SOURCE_DATA_DIR('train.zip', external_data_url)
source_data_dir = hutton_v1_dataset.get_SOURCE_DATA_DIR()
# Set the active data source directory
active_data_dir = data_dir
hutton_v1_dataset.set_ACTIVE_DATA_DIR(active_data_dir)
# Prepare the training and validation datasets
hutton_v1_dataset.prepare_train_dataset()
hutton_v1_dataset.prepare_validation_dataset()
# Create batches
image_batch, label_batch = hutton_v1_dataset.create_batches()

• [Optional] Set up the parameters

img_height = hutton_v1_dataset.get_IMG_HEIGHT()
img_width = hutton_v1_dataset.get_IMG_WIDTH()
batch_size = hutton_v1_dataset.get_BATCH_SIZE()
number_classes = hutton_v1_dataset.get_NUM_CLASSES()
current_train_ds = hutton_v1_dataset.get_train_dataset()
current_val_ds = hutton_v1_dataset.get_validation_dataset()
first_image = image_batch[0]