pp6hrvtlm/cnn_for_license_plate_recognition/process_data.py

import os
import random
from PIL import Image


def image2matrix(image_filename):
    image = Image.open(image_filename)
    image = image.resize((20, 20))
    matrix = [[image.getpixel((x, y)) for x in range(0, image.size[0])] for y in range(0, image.size[1])]
    return matrix


class DataProcessor:
    def __init__(self, aspect='area', separate_ratio=0.1):
        self.train_dir = './dataset/train/{}/'.format(aspect)
        self.validate_dir = './dataset/val/{}/'.format(aspect)
        self.separate_ratio = separate_ratio

        self.vectors = []
        self.labels = []

        self.train_set = []
        self.train_batch_index = 0
        self.train_epoch = 0
        self.validate_set = []
        self.validate_batch_index = 0
        self.test_set = []
        self.test_batch_index = 0

        self.classes = 0
        self.data_set_count = 0

        self.load_train()
        self.load_valid()

    def load_train(self):
        for root, dirs, files in os.walk(self.train_dir):
            self.classes = max(self.classes, len(dirs))
            if len(dirs) == 0:
                label = int(root.split('/')[-1])
                for name in files:
                    image_filename = os.path.join(root, name)
                    vector = image2matrix(image_filename)
                    self.vectors.append(vector)
                    self.labels.append(label)
                    if random.random() < self.separate_ratio:
                        self.test_set.append(self.data_set_count)
                    else:
                        self.train_set.append(self.data_set_count)
                    self.data_set_count += 1

    def load_valid(self):
        for root, dirs, files in os.walk(self.validate_dir):
            self.classes = max(self.classes, len(dirs))
            if len(dirs) == 0:
                label = int(root.split('/')[-1])
                for name in files:
                    image_filename = os.path.join(root, name)
                    vector = image2matrix(image_filename)
                    self.vectors.append(vector)
                    self.labels.append(label)
                    if random.random() < self.separate_ratio:
                        self.test_set.append(self.data_set_count)
                    else:
                        self.validate_set.append(self.data_set_count)
                    self.data_set_count += 1

    def next_train_batch(self, batch=100):
        input_x = []
        input_y = []
        for i in range(batch):
            input_x.append(self.vectors[self.train_set[(self.train_batch_index + i) % len(self.train_set)]])
            y = [0] * 34
            y[self.labels[self.train_set[(self.train_batch_index + i) % len(self.train_set)]]] = 1
            input_y.append(y)
        self.train_batch_index += batch
        if self.train_batch_index > len(self.train_set):
            self.train_epoch += 1
        self.train_batch_index %= len(self.train_set)
        return input_x, input_y, self.train_epoch

    def next_valid_batch(self, batch=100):
        input_x = []
        input_y = []
        for i in range(batch):
            index = random.randint(0, len(self.validate_set) - 1)
            input_x.append(self.vectors[index])
            y = [0] * 34
            y[self.labels[index]] = 1
            input_y.append(y)
        self.validate_batch_index += batch
        self.validate_batch_index %= len(self.validate_set)
        return input_x, input_y, self.train_epoch

    def next_test_batch(self, batch=100):
        input_x = []
        input_y = []
        for i in range(batch):
            input_x.append(self.vectors[self.test_set[(self.test_batch_index + i) % len(self.test_set)]])
            y = [0] * 34
            y[self.labels[self.test_set[(self.test_batch_index + i) % (len(self.test_set))]]] = 1
            input_y.append(y)
        self.test_batch_index += batch
        if self.test_batch_index > len(self.test_set):
            self.train_epoch += 1
        self.test_batch_index %= len(self.test_set)
        return input_x, input_y, self.train_epoch