Hands-On Python Deep Learning for the Web

Creating functions to read label files

There are two label files available to us in the MNIST dataset: train-labels-idx1-ubyte and t10k-labels-idx1-ubyte. To view these files, we can use the following function, which takes input of the filename as an argument and produces an array of one-hot-encoded labels:

def loadLabelFile(filelabel):  f = open(filelabel, "rb")  f.read(8)  labels_arr = []  while True:    row = [0 for x in range(10)]    try:      label = ord(f.read(1))      row[label] = 1      labels_arr.append(row)    except:      break  f.close()  label_sets = np.array(labels_arr)  return label_sets

This function returns a numpy array of labels in one-hot encoding, with the dimensions of the number of samples in the dataset times by 10. Let's observe a single entry in order ...