Skip to content

Commit

Permalink
add cn doc for RandomPerspective and perspective (#5353)
Browse files Browse the repository at this point in the history
  • Loading branch information
nemonameless authored Oct 20, 2022
1 parent 8f27bb5 commit ecfd3a0
Show file tree
Hide file tree
Showing 2 changed files with 77 additions and 0 deletions.
42 changes: 42 additions & 0 deletions docs/api/paddle/vision/transforms/RandomPerspective_cn.rst
Original file line number Diff line number Diff line change
@@ -0,0 +1,42 @@
.. _cn_api_vision_transforms_RandomPerspective:

RandomPerspective
-------------------------------

.. py:class:: paddle.vision.transforms.RandomPerspective(prob=0.5, distortion_scale=0.5, interpolation='nearest', fill=0, keys=None)
按照一定概率对图片进行透视变换。

参数
::::::::::::
- **prob** (float,可选) - 进行透视变换的概率,范围为 [0, 1] 。默认值: 0.5 。
- **distortion_scale** (float,可选) - 图片失真程度的大小,范围为 [0, 1] 。默认值: 0.5 。
- **interpolation** (str,可选) - 插值的方法。
如果这个参数没有设定或者输入图像为单通道,则该参数会根据使用的后端,被设置为 ``PIL.Image.NEAREST`` 或者 ``cv2.INTER_NEAREST`` 。
当使用 ``pil`` 作为后端时, 支持的插值方法如下:
- "nearest": Image.NEAREST
- "bilinear": Image.BILINEAR
- "bicubic": Image.BICUBIC
当使用 ``cv2`` 作为后端时, 支持的插值方法如下:
- "nearest": cv2.INTER_NEAREST
- "bilinear": cv2.INTER_LINEAR
- "bicubic": cv2.INTER_CUBIC
- **fill** (int|list|tuple,可选) - 对图像扩展时填充的值。默认值: 0 ,如果只设定一个数字则所有通道上像素值均为该值。
- **keys** (list[str]|tuple[str],可选) - 与 ``BaseTransform`` 定义一致。默认值: None 。

形状
::::::::::::

- img (PIL.Image|np.ndarray|Paddle.Tensor) - 输入的图像数据,数据格式为 [H, W, C] 。
- output (PIL.Image|np.ndarray|Paddle.Tensor) - 返回随机透视变换后的图像数据。

返回
::::::::::::

计算 ``RandomPerspective`` 的可调用对象。


代码示例
::::::::::::

COPY-FROM: paddle.vision.transforms.RandomPerspective
35 changes: 35 additions & 0 deletions docs/api/paddle/vision/transforms/perspective_cn.rst
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
.. _cn_api_vision_transforms_perspective:

perspective
-------------------------------

.. py:function:: paddle.vision.transforms.perspective(img, startpoints, endpoints, interpolation='nearest', fill=0)
对图像进行透视变换。

参数
::::::::::::
- **img** (PIL.Image|numpy.ndarray|paddle.Tensor) - 输入图像。
- **startpoints** (list(list(float))) - 在原图上的四个角(左上、右上、右下、左下)的坐标。
- **endpoints** (list(list(float))) - 在变换后的图上的四个角(左上、右上、右下、左下)的坐标。
- **interpolation** (str,可选) - 插值的方法。
如果这个参数没有设定或者输入图像为单通道,则该参数会根据使用的后端,被设置为 ``PIL.Image.NEAREST`` 或者 ``cv2.INTER_NEAREST`` 。
当使用 ``pil`` 作为后端时, 支持的插值方法如下:
- "nearest": Image.NEAREST
- "bilinear": Image.BILINEAR
- "bicubic": Image.BICUBIC
当使用 ``cv2`` 作为后端时, 支持的插值方法如下:
- "nearest": cv2.INTER_NEAREST
- "bilinear": cv2.INTER_LINEAR
- "bicubic": cv2.INTER_CUBIC
- **fill** (int|list|tuple,可选) - 对图像扩展时填充的像素值,默认值: 0 ,如果只设定一个数字则所有通道上像素值均为该值。

返回
::::::::::::

``PIL.Image 或 numpy ndarray 或 paddle.Tensor`` ,透视变换后的图像。

代码示例
::::::::::::

COPY-FROM: paddle.vision.transforms.perspective

0 comments on commit ecfd3a0

Please sign in to comment.