將MNIST手寫數據集轉換成圖片保存到本地

# 加載圖片
data = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = data.load_data()

plt.imshow(x_train[0], cmap='gray')

# 創建10個文件夾存放每一類圖片
for i in range(10):
    os.makedirs(f"../datas/mnist/train/{i}")
    os.makedirs(f"../datas/mnist/test/{i}")

# 保存圖片並生成圖像列表

# 訓練集數據
with open("../datas/mnist/train/image_list.txt", 'w') as img_list:
    i = 1
    for img, label in zip(x_train, y_train):
        img = Image.fromarray(img) # 將array轉化成圖片
        img_save_path = f"../datas/mnist/train/{label}/{i}.jpg" # 圖片保存路徑
        img.save(img_save_path) # 保存圖片
        img_list.write(img_save_path + "\t" + str(label) + "\n")
        i += 1

# 測試集數據
with open("../datas/mnist/test/image_list.txt", 'w') as img_list:
    i = 1
    for img, label in zip(x_test, y_test):
        img = Image.fromarray(img) # 將array轉化成圖片
        img_save_path = f"../datas/mnist/test/{label}/{i}.jpg" # 圖片保存路徑
        img.save(img_save_path)
        img_list.write(img_save_path + "\t" + str(label) + "\n")
        i += 1