[paddleseg source code reading] normalize operation of paddleseg transform

Connected to a ,PaddleSeg Custom data class https://blog.csdn.net/HaoZiHuang/article/details/125566058 Let's talk about it here Normalize operation

Let's write it ourselves Transform when , Always worry about some details , such as ：

• np.uint8 Did you turn , Do you want to change to np.float32
• Have you done it yet (/255 Did you do it )
• With or without subtracting 0.5, Then divide by 0.5, Move it to -1 To 1
• Before sending data , Is there any transpose And so on.
• When will numpy.ndarray Initialize to paddle.Tensor

Last one PaddleSeg Speaking of , stay Compose ending , Can convert the number of input channels (Transpose), And look at the source code ：

batch_sampler = paddle.io.DistributedBatchSampler(
        eval_dataset, batch_size=1, shuffle=False, drop_last=False)
loader = paddle.io.DataLoader(
    eval_dataset,
    batch_sampler=batch_sampler,
    num_workers=num_workers,
    return_list=True, )

This is instantiation ,Dataset and DataLoader Part of , Even though eval_dataset Every data returned is np.ndarray, however paddle.io.DataLoader This class , Will go straight back to Paddle.Tensor Class , That is, there is no need to do paddle.to_tensor operation

OK, The next three items ：

• np.uint8 Did you turn , Do you want to change to np.float32
• Have you done it yet
• With or without subtracting 0.5, Then divide by 0.5, Move it to -1 To 1

Namely Normalize Part of the operation , Take a look at the source code of its implementation

paddleseg\transforms\transforms.py in Normalize class

@manager.TRANSFORMS.add_component
class Normalize:
    """ Normalize an image. Args: mean (list, optional): The mean value of a data set. Default: [0.5, 0.5, 0.5]. std (list, optional): The standard deviation of a data set. Default: [0.5, 0.5, 0.5]. Raises: ValueError: When mean/std is not list or any value in std is 0. """

    def __init__(self, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5)):
        self.mean = mean
        self.std = std
        if not (isinstance(self.mean,
                           (list, tuple)) and isinstance(self.std,
                                                         (list, tuple))):
            raise ValueError(
                "{}: input type is invalid. It should be list or tuple".format(
                    self))
        from functools import reduce
        if reduce(lambda x, y: x * y, self.std) == 0:
            raise ValueError('{}: std is invalid!'.format(self))

    def __call__(self, im, label=None):
        """ Args: im (np.ndarray): The Image data. label (np.ndarray, optional): The label data. Default: None. Returns: (tuple). When label is None, it returns (im, ), otherwise it returns (im, label). """

        mean = np.array(self.mean)[np.newaxis, np.newaxis, :]
        std = np.array(self.std)[np.newaxis, np.newaxis, :]
        im = functional.normalize(im, mean, std)

        if label is None:
            return (im, )
        else:
            return (im, label)

paddleseg\transforms\functional.py in normalize function

def normalize(im, mean, std):
    im = im.astype(np.float32, copy=False) / 255.0
    im -= mean
    im /= std
    return im

You can see stay normalize Function , Made a turn np.float32, normalization

then -0.5 after /0.5

---

If you are interested in these two fields ：

mean = np.array(self.mean)[np.newaxis, np.newaxis, :]
std  = np.array(self.std )[np.newaxis, np.newaxis, :]

Print it yourself , It's the original array The dimension is preceded by two 1
shape from (3,) Turned into (1, 1, 3)

and np.newaxis In fact, that is None

(np.newaxis == None) == True

---

If the Normalize __init__ Function reduce Interested in ：
Look at this ：
functools Under the reduce function