utils.py

#!/usr/bin/env python3
"""
Utility Script containing functions to be used for training
Author: Shilpaj Bhalerao
"""
# Standard Library Imports
import math
from typing import NoReturn
import io
from PIL import Image
# Third-Party Imports
import numpy as np
import matplotlib.pyplot as plt
import torch
from torchsummary import summary
from torchvision import transforms
from pytorch_grad_cam import GradCAM
from pytorch_grad_cam.utils.image import show_cam_on_image


def get_summary(model, input_size: tuple) -> NoReturn:
    """
    Function to get the summary of the model architecture
    :param model: Object of model architecture class
    :param input_size: Input data shape (Channels, Height, Width)
    """
    use_cuda = torch.cuda.is_available()
    device = torch.device("cuda" if use_cuda else "cpu")
    network = model.to(device)
    summary(network, input_size=input_size)


def get_misclassified_data(model, device, test_loader):
    """
    Function to run the model on test set and return misclassified images
    :param model: Network Architecture
    :param device: CPU/GPU
    :param test_loader: DataLoader for test set
    """
    # Prepare the model for evaluation i.e. drop the dropout layer
    model.eval()

    # List to store misclassified Images
    misclassified_data = []

    # Reset the gradients
    with torch.no_grad():
        # Extract images, labels in a batch
        for data, target in test_loader:

            # Migrate the data to the device
            data, target = data.to(device), target.to(device)

            # Extract single image, label from the batch
            for image, label in zip(data, target):

                # Add batch dimension to the image
                image = image.unsqueeze(0)

                # Get the model prediction on the image
                output = model(image)

                # Convert the output from one-hot encoding to a value
                pred = output.argmax(dim=1, keepdim=True)

                # If prediction is incorrect, append the data
                if pred != label:
                    misclassified_data.append((image, label, pred))
    return misclassified_data


# -------------------- DATA STATISTICS --------------------
def get_mnist_statistics(data_set, data_set_type='Train'):
    """
    Function to return the statistics of the training data
    :param data_set: Training dataset
    :param data_set_type: Type of dataset [Train/Test/Val]
    """
    # We'd need to convert it into Numpy! Remember above we have converted it into tensors already
    train_data = data_set.train_data
    train_data = data_set.transform(train_data.numpy())

    print(f'[{data_set_type}]')
    print(' - Numpy Shape:', data_set.train_data.cpu().numpy().shape)
    print(' - Tensor Shape:', data_set.train_data.size())
    print(' - min:', torch.min(train_data))
    print(' - max:', torch.max(train_data))
    print(' - mean:', torch.mean(train_data))
    print(' - std:', torch.std(train_data))
    print(' - var:', torch.var(train_data))

    dataiter = next(iter(data_set))
    images, labels = dataiter[0], dataiter[1]

    print(images.shape)
    print(labels)

    # Let's visualize some of the images
    plt.imshow(images[0].numpy().squeeze(), cmap='gray')


def get_cifar_property(images, operation):
    """
    Get the property on each channel of the CIFAR
    :param images: Get the property value on the images
    :param operation: Mean, std, Variance, etc
    """
    param_r = eval('images[:, 0, :, :].' + operation + '()')
    param_g = eval('images[:, 1, :, :].' + operation + '()')
    param_b = eval('images[:, 2, :, :].' + operation + '()')
    return param_r, param_g, param_b


def get_cifar_statistics(data_set, data_set_type='Train'):
    """
    Function to get the statistical information of the CIFAR dataset
    :param data_set: Training set of CIFAR
    :param data_set_type: Training or Test data
    """
    # Images in the dataset
    images = [item[0] for item in data_set]
    images = torch.stack(images, dim=0).numpy()

    # Calculate mean over each channel
    mean_r, mean_g, mean_b = get_cifar_property(images, 'mean')

    # Calculate Standard deviation over each channel
    std_r, std_g, std_b = get_cifar_property(images, 'std')

    # Calculate min value over each channel
    min_r, min_g, min_b = get_cifar_property(images, 'min')

    # Calculate max value over each channel
    max_r, max_g, max_b = get_cifar_property(images, 'max')

    # Calculate variance value over each channel
    var_r, var_g, var_b = get_cifar_property(images, 'var')

    print(f'[{data_set_type}]')
    print(f' - Total {data_set_type} Images: {len(data_set)}')
    print(f' - Tensor Shape: {images[0].shape}')
    print(f' - min: {min_r, min_g, min_b}')
    print(f' - max: {max_r, max_g, max_b}')
    print(f' - mean: {mean_r, mean_g, mean_b}')
    print(f' - std: {std_r, std_g, std_b}')
    print(f' - var: {var_r, var_g, var_b}')

    # Let's visualize some of the images
    plt.imshow(np.transpose(images[1].squeeze(), (1, 2, 0)))


# -------------------- GradCam --------------------
def display_gradcam_output(data: list,
                           classes,
                           inv_normalize: transforms.Normalize,
                           model,
                           target_layers,
                           targets=None,
                           number_of_samples: int = 10,
                           transparency: float = 0.60):
    """
    Function to visualize GradCam output on the data
    :param data: List[Tuple(image, label)]
    :param classes: Name of classes in the dataset
    :param inv_normalize: Mean and Standard deviation values of the dataset
    :param model: Model architecture
    :param target_layers: Layers on which GradCam should be executed
    :param targets: Classes to be focused on for GradCam
    :param number_of_samples: Number of images to print
    :param transparency: Weight of Normal image when mixed with activations
    """
    # Plot configuration
    fig = plt.figure(figsize=(10, 10))
    x_count = 5
    y_count = math.ceil(number_of_samples / x_count) 

    # Create an object for GradCam
    cam = GradCAM(model=model, target_layers=target_layers)

    # Iterate over number of specified images
    for i in range(number_of_samples):
        plt.subplot(y_count, x_count, i + 1)
        input_tensor = data[i][0]

        # Get the activations of the layer for the images
        grayscale_cam = cam(input_tensor=input_tensor, targets=targets)
        grayscale_cam = grayscale_cam[0, :]

        # Get back the original image
        img = input_tensor.squeeze(0).to('cpu')
        img = inv_normalize(img)
        rgb_img = np.transpose(img, (1, 2, 0))
        rgb_img = rgb_img.numpy().astype(np.float32)

# Ensure the image data is within the [0, 1] range
        rgb_img = np.clip(rgb_img, 0, 1)

        # Mix the activations on the original image
        visualization = show_cam_on_image(rgb_img, grayscale_cam, use_rgb=True, image_weight=transparency)

        # Display the images on the plot
        plt.imshow(visualization)
        plt.title(r"Correct: " + classes[data[i][1].item()] + '\n' + 'Output: ' + classes[data[i][2].item()])
        plt.xticks([])
        plt.yticks([])

    plt.tight_layout()

    # Save the entire figure to a BytesIO object
    buf = io.BytesIO()
    plt.savefig(buf, format='png')
    buf.seek(0)
    img_var = Image.open(buf)

    return img_var