mlplot.py

from matplotlib import pyplot as plt
import numpy as np
import seaborn as sns
import pandas as pd
from PIL import Image

def plot_gan_samples(samples):
    fig, axes = plt.subplots(figsize=(10, 2), nrows=1, ncols=len(samples), sharey=True, sharex=True)
    for ax, img in zip(axes.flatten(), samples):
        ax.xaxis.set_visible(False)
        ax.yaxis.set_visible(False)
        im = ax.imshow(1 - img.reshape((2,2)), cmap='Greys_r')  
    return fig, axes

def plot_generator(samples, epoch):
    fig, axes = plt.subplots(figsize=(10, 2), nrows=1, ncols=len(samples), sharey=True, sharex=True)
    fig.suptitle(f'epoch: {epoch}', ha='left', fontweight='light')
    for ax, img in zip(axes.flatten(), samples):
        ax.xaxis.set_visible(False)
        ax.yaxis.set_visible(False)
        im = ax.imshow(1 - img.reshape((2,2)), cmap='Greys_r')  
        plt.close(fig)
    return fig, axes

def fig2img(fig):
    """Convert a Matplotlib figure to a PIL Image and return it"""
    import io
    buf = io.BytesIO()
    fig.savefig(buf)
    buf.seek(0)
    img = Image.open(buf)
    return img

def plot_log():
    x = np.linspace(0.001, 0.999, 100)
    E_r = lambda x: -np.log(x)
    plt.plot(x, E_r(x), label=r'$- \, log \, (x)$')
    E_f = lambda x: -np.log(1 - x)
    plt.plot(x, E_f(x), label=r'$- \, log \, (1 - x)$')
    plt.legend(fontsize='large')
    plt.xlabel('Prediction')
    plt.ylabel('Error')
    return True

def plot_errors(generator_errors, discriminator_errors):
    plt.figure(figsize=(16, 5))
    plt.subplot(1,2,1)  
    plt.plot(generator_errors)
    plt.title("Generator Error")
    plt.xlabel("Epoche")
    plt.ylabel("Error")
    plt.subplot(1,2,2)
    plt.plot(discriminator_errors)
    plt.title("Discriminator Error")
    plt.xlabel("Epoche")
    plt.ylabel("Error")
    plt.show()
    
def plot_discriminator_accuracy(accs_real, accs_fake):
    plt.figure(figsize=(16, 5))
    plt.subplot(1,2,1)  
    plt.plot(accs_real)
    plt.title("Real Data Accuracy")
    plt.xlabel("Epoche")
    plt.ylabel("Accuracy")
    plt.subplot(1,2,2)
    plt.plot(accs_fake)
    plt.title("Fake Data Accuracy")
    plt.xlabel("Epoche")
    plt.ylabel("Accuracy")
    plt.show()
    
def plot_1d_samples(generate_real_samples, generate_fake_samples, n=100):
    Xr, yr = generate_real_samples(n)
    dfr, dfr['real'] = pd.DataFrame(Xr, columns=['input', 'output']), yr
    Xf, yf = generate_fake_samples(n)
    dff, dff['real'] = pd.DataFrame(Xf, columns=['input', 'output']), yf
    df = pd.concat([dfr, dff], ignore_index=True)
    sns.scatterplot(data=df, x='input', y='output', hue='real')
    
def plot_1d_performance(epoch, x_real, x_fake, accuracy_real, accuracy_fake):
    fig = plt.figure(figsize=(19, 5))
    plt.subplot(1,3,1)
    plt.xlim(-1.25, 1.25)
    plt.ylim(-0.25, 0.75)
    plt.title(f'Epoch: {epoch:0>5d}')
    plt.scatter(x_real[:, 0], x_real[:, 1], color='red', alpha=0.3)
    plt.scatter(x_fake[:, 0], x_fake[:, 1], color='blue', alpha=0.3)
    plt.subplot(1,3,2)
    plt.title(f'Accuracy Real: {accuracy_real[-1]:.3f}')
    plt.plot(accuracy_real)
    plt.subplot(1,3,3)
    plt.title(f'Accuracy Fake: {accuracy_fake[-1]:.3f}')
    plt.plot(accuracy_fake)
    plt.close(fig)
    return fig