get_dataset.py 文件源码

def get_dataset(dataset_path='Data/Train_Data'):
    # Getting all data from data path:
    try:
        X = np.load('Data/npy_train_data/X.npy')
        Y = np.load('Data/npy_train_data/Y.npy')
    except:
        inputs_path = dataset_path+'/input'
        images = listdir(inputs_path) # Geting images
        X = []
        Y = []
        for img in images:
            img_path = inputs_path+'/'+img

            x_img = get_img(img_path).astype('float32').reshape(64, 64, 3)
            x_img /= 255.

            y_img = get_img(img_path.replace('input/', 'mask/mask_')).astype('float32').reshape(64, 64, 1)
            y_img /= 255.

            X.append(x_img)
            Y.append(y_img)
        X = np.array(X)
        Y = np.array(Y)
        # Create dateset:
        if not os.path.exists('Data/npy_train_data/'):
            os.makedirs('Data/npy_train_data/')
        np.save('Data/npy_train_data/X.npy', X)
        np.save('Data/npy_train_data/Y.npy', Y)
    X, X_test, Y, Y_test = train_test_split(X, Y, test_size=0.1, random_state=42)
    return X, X_test, Y, Y_test