caffe python接口图片预处理 transformer 类学习笔记

最新推荐文章于 2024-08-17 18:22:32 发布

zuqiutxy

最新推荐文章于 2024-08-17 18:22:32 发布

阅读量2.4k

点赞数

本文链接：https://blog.csdn.net/zuqiutxy/article/details/52958132

版权

本文介绍了Caffe库中Transformer类的使用，包括设置transpose、channel_swap、raw_scale、mean等参数，以实现图片从RGB到BGR的转换、尺寸调整、归一化等预处理步骤，便于输入到神经网络模型。

摘要由CSDN通过智能技术生成

1:调用该接口, 传入字典参数
eg: transformer=caffe.io.Transformer({'data': net.blobs['data'].data.shape})

2: self.set_transpose(self,in_,order) 用于将输入图片通道进行重排，将RGB格式转换为BGR
in_ 字典的key值
order: 图像通道重排的顺序
eg:transformer.set_transpose('data', (2,0,1))

3：self.set_channel_swap(self, in_, order)
用于将输入图片通道进行重排，将RGB格式转换为BGR
in_ 字典的key值
order: 图像通道重排的顺序

class Transformer:
    """
    Transform input for feeding into a Net.

    Note: this is mostly for illustrative purposes and it is likely better
    to define your own input preprocessing routine for your needs.

    Parameters
    ----------
    net : a Net for which the input should be prepared
    """
    def __init__(self, inputs):    #inputs :输入字典 eg:{'data': net.blobs['data'].data.shape}
        self.inputs = inputs
        self.transpose = {}
        self.channel_swap = {}
        self.raw_scale = {}
        self.mean = {}
        self.input_scale = {}

    def __check_input(self, in_):          # in_: 字典的key值
        if in_ not in self.inputs:
            raise Exception('{}