[输入参数类型不一致]torchvision.transforms.RandomPerspective

torchvision.transforms.RandomPerspective

torchvision.transforms.RandomPerspective(
    distortion_scale: float = 0.5,
    p: float = 0.5,
    interpolation: InterpolationMode = InterpolationMode.BILINEAR,
    fill: Union[int, float, List[int], Tuple[int, ...]] = 0
)

paddle.vision.transforms.RandomPerspective

paddle.vision.transforms.RandomPerspective(
    prob: float = 0.5,
    distortion_scale: float = 0.5,
    interpolation: Union[str, int] = 'nearest',
    fill: Union[int, List[int], Tuple[int, ...]] = 0,
    keys: Optional[Union[List[str], Tuple[str, ...]]] = None
)

两者功能一致,但参数类型不一致。

参数映射

torchvision PaddlePaddle 备注
distortion_scale distortion_scale 控制失真程度。
p prob 进行透视变换的概率,仅参数名不一致。
interpolation interpolation 插值的方法,两者类型不一致,PyTorch 为 InterpolationMode 枚举类, Paddle 为 int 或 string,需要转写。
fill fill 对图像扩展时填充的值。
- keys 输入的类型,PyTorch 无此参数,Paddle 保持默认即可。

转写示例

interpolation:插值的方法

# PyTorch 写法
transform = torchvision.transforms.RandomPerspective(distortion_scale=0.5, p=0.5, interpolation=torchvision.transforms.InterpolationMode.BILINEAR, fill=0)
transformed_img = transform(img)

# Paddle 写法
transform = paddle.vision.transforms.RandomPerspective(distortion_scale=0.5, prob=0.5, interpolation='bilinear', fill=0)
transformed_img = transform(img)