multi-scale-expansion 0.1.2

Check whether your plants are healthy using image classification

Multi Scale Expansion

Repository to create framework for image classification computer vision models in tensorflow.

Overview

multi-scale-expansion is a library for automating the set up of an image classification model. The user provides their data, and the library creates and trains a ready-to-use model to complete the image classification task and apply it to any image further. The objective is that this framework can be automated and applied for recognizing whether a plant is healthy or not, through the use of the models we train.

Contributions

For instructions on how to contribute, go to the Contribution Guidelines Page.

Installation

Prerequisites:

• Python >= 3.7
• Torch & Torchvision
• Numpy
• Matplotlib
• PIL (Pillow)

To install Python packages:

$ pip install torch
$ pip install torchvision
$ pip install numpy
$ pip install matplotlib
$ pip install Pillow

To install library:

$ pip install multi-scale-expansion

Quick-Start Example

# Provide a pytorch pre-trained image classification model! 
mock_model = ms_model.get_plant_model(mock_model, list(range(6))

# Specify these values - fine-tune at your will!
mock_criterion, mock_optimizer, mock_lr_scheduler = get_train_loss_needs(
    mock_model, mock_lr, mock_momentum, mock_step_size, mock_gamma
)

# Get dataloaders from datasets
mock_dataloaders = ms_datasets.get_dataloaders(mock_datasets)

# Fine-tune your pre-trained model! 
model, train_losses, train_accuracies, val_losses, val_accuracies = ms.train_model(
    device,
    mock_dataset_sizes,
    mock_dataloaders,
    mock_model,
    mock_criterion,
    mock_optimizer,
    mock_lr_scheduler,
    num_epochs=1,
    testing=True,
)

And now your model is ready-to-use for your image classification task! Soon, you'll be able to just call a single method and the library will set up the whole classification task for you!