prepare_data.py

import numpy as np
import os
from PIL import Image


def load(dir, files, reshaped):
    "Load .npy or .npz files from disk and return them as numpy arrays. \
    Takes in a list of filenames and returns a list of numpy arrays."

    data = []
    for file in files:
        f = np.load(dir + file)
        if reshaped:
            new_f = []
            for i in range(len(f)):
                x = np.reshape(f[i], (28, 28))
                x = np.expand_dims(x, axis=0)
                x = np.reshape(f[i], (28, 28, 1))
                new_f.append(x)
            f = new_f
        data.append(f)
    return data


def normalize(data):
    "Takes a list or a list of lists and returns its normalized form"

    return np.interp(data, [0, 255], [-1, 1])


def denormalize(data):
    "Takes a list or a list of lists and returns its denormalized form"

    return np.interp(data, [-1, 1], [0, 255])


def visualize(array):
    "Visulaze a 2D array as an Image"

    img = Image.fromarray(array)
    img.show(title="Visulizing array")


def set_limit(arrays, n):
    "Limit elements from each array up to n elements and return a single list"
    new = []
    for array in arrays:
        i = 0
        for item in array:
            if i == n:
                break
            new.append(item)
            i += 1
    return new


def make_labels(N1, N2):
    "make labels from 0 to N1, each repeated N2 times"
    labels = []
    for i in range(N1):
        labels += [i] * N2
    return labels