NeuroClassify Usage Guide

NeuroClassify is a Python package designed to help you easily create, train, save, and use image classification models with TensorFlow and Keras. The package provides functionality for training a model on a dataset, saving and loading the model, and making predictions on images.

Table of Contents

• Installation
• Usage
  • Creating an Image Classifier
  • Training the Model
  • Saving and Loading the Model
  • Making Predictions
  • Predicting Multiple Images
• Functions

---

Installation

To install the NeuroClassify package, you can use pip:

pip install neuroclassify

---

Usage

Creating an Image Classifier

To create an image classifier, you need to specify the base directory where your training images are stored. The images should be organized into subdirectories named after the class labels. Here's how you can set up the classifier:

from neuroclassify import ImageClassifier

# Create an instance of the ImageClassifier
classifier = ImageClassifier(img_size=(150, 150), batch_size=32)
classifier.dataset_dir = 'path_to_your_dataset'

In the above code, path_to_your_dataset should be the directory containing subdirectories for training and validation images.

Training the Model

After creating the classifier, you can train it by calling the train method. You can specify the number of epochs for training:

history = classifier.train(epochs=20)  # Train the model for 20 epochs

The method will automatically load the images from the train and val subdirectories inside your dataset directory and start the training process.

---

Saving and Loading the Model

You can save the trained model to a file and load it back later using the following methods:

Save Model

To save the model, call the save_model method. This will save both the model in a .h5 file and the class labels in a labels.txt file:

classifier.save_model(name='my_model')  # Save with a custom filename

This will also create a .zip file that includes the model and the labels.txt file.

Load Model

To load the model, use the load_model method, providing the paths to the saved model and label files:

classifier.load_model(model='my_model.h5', label='labels.txt')  # Load the model from the specified file

---

Making Predictions

To make predictions with the trained model, you can use the predict_image() function. This allows you to classify a single image and get the predicted class name.

Predicting a Single Image

Here is how you can predict the class of a single image:

# Define the path to your model, image, and labels
img_path = 'path_to_your_image/image.jpg'
classifier.load_model(model='my_model.h5', label='labels.txt')  # Load the model from the specified file

# Predict the class of an image and display the image
predicted_class_name = predict_image(img_path, display=True)
print(f'Predicted Class: {predicted_class_name}')

The display=True option will show the image along with the predicted class name using Matplotlib.

Predicting Multiple Images

To predict the classes of all images in a directory, you can use the predict_images function. This will return a list of predictions for each image:

# Predict classes for all images in a directory
img_dir = 'path_to_your_image_directory'  # Directory containing images
predictions = predict_images(img_dir, display=True)

for filename, predicted_class_name in predictions:
    print(f'File: {filename}, Predicted Class: {predicted_class_name}')

This will go through all images in the specified directory and print the predictions for each one.

---

Functions

ImageClassifier Class

• create_model: Builds and compiles the image classification model.
• train: Trains the model using images from the dataset.
• save_model: Saves the trained model and its labels.
• load_model: Loads a pre-trained model and its labels.

predict_image

Predicts the class of a single image and optionally displays it.

• Arguments:
  • model: The trained model to use for predictions.
  • img_path: Path to the image file to predict.
  • label_file: Path to the label file.
  • display (optional): If True, displays the image with its predicted class.

predict_images

Predicts the classes for all images in a directory and optionally displays them.

• Arguments:
  • model: The trained model to use for predictions.
  • img_dir: Directory containing images to predict.
  • label_file: Path to the label file.
  • display (optional): If True, displays images with their predicted classes.

load_labels

Loads class labels from a text file (used internally for predictions).

---

Contributing

Contributions to the NeuroClassify package are welcome! If you have suggestions or improvements, feel free to submit a pull request.

---

License

NeuroClassify is open source and available under the MIT License.

---

This guide provides everything you need to get started with NeuroClassify for image classification. You can now train models, save/load them, and use them for predictions with ease.

---