Created by: yeyupiaoling

环境

• PaddlePaddle 1.2.0
• Python 3.5
• Ubuntu 16.04
• 模型：MobileNet V1

问题

我使用以下的Python代码读取图像数据，但是每一次价值数据都很慢，比训练时间还长，这是什么原因呢？是否能解决？

# 训练图片的预处理
def train_mapper(sample):
    img, label = sample
    path = img
    img = Image.open(img)
    # 统一图片大小
    try:
        img = img.resize((cfg.READER.RESIZE_SIZE, cfg.READER.RESIZE_SIZE), Image.ANTIALIAS)
    except:
        print(path)
    # 随机水平翻转
    r1 = random.random()
    if r1 > 0.5:
        img = img.transpose(Image.FLIP_LEFT_RIGHT)
    # 随机垂直翻转
    r2 = random.random()
    if r2 > 0.5:
        img = img.transpose(Image.FLIP_TOP_BOTTOM)
    # 随机角度翻转
    r3 = random.randint(-3, 3)
    img = img.rotate(r3, expand=False)
    # 随机裁剪
    r4 = random.randint(0, int(cfg.READER.RESIZE_SIZE - cfg.READER.CROP_SIZE))
    r5 = random.randint(0, int(cfg.READER.RESIZE_SIZE - cfg.READER.CROP_SIZE))
    box = (r4, r5, r4 + cfg.READER.CROP_SIZE, r5 + cfg.READER.CROP_SIZE)
    img = img.crop(box)
    # 把图片转换成numpy值
    img = np.array(img).astype(np.float32)
    # 转换成CHW
    img = img.transpose((2, 0, 1))
    # 转换成BGR
    img = img[(2, 1, 0), :, :] / 255.0

    return img, int(label)


# 获取训练的reader
def train_reader(train_list_path):
    father_path = os.path.dirname(train_list_path)

    def reader():
        with open(train_list_path, 'r') as f:
            lines = f.readlines()
            for line in lines:
                img, label = line.split('\t')
                img = os.path.join(father_path, img)
                yield img, label

    return paddle.reader.xmap_readers(train_mapper, reader, cpu_count(), 1024)