【Pytorch】自定义的transfomer模块(torchvision)---ZeroPaddingResize

1.目的

有时候直接进行resize会有形变，所以想到这样的方式，同比例缩放，然后补0。

2.实现

参考：https://www.pythonf.cn/read/144858

class ZeroPaddingResize(object):
    """
        resize image with same scale, fill with zero padding
    """
    def __init__(self, size=(224, 224), interpolation=Image.BILINEAR):
        self.size = size
        self.interpolation = interpolation

    def __call__(self, img):
        ih, iw = img.size
        h, w = self.size[0], self.size[1]
        scale = min(w / iw, h / ih)
        new_w = int(iw * scale+0.5)
        new_h = int(ih * scale+0.5)

        img = img.resize((new_w, new_h), self.interpolation)
        new_img = Image.new('RGB', self.size, (0, 0, 0))
        new_img.paste(img, ((w-new_w)//2, (h-new_h)//2))

        return new_img

3.使用

import torchvision
import ZeroPaddingResize

transform_train = torchvision.transforms.Compose([
    ZeroPaddingResize((224, 224)),
    torchvision.transforms.ToTensor(),
])