image_transform

Image transform abstractions and helpers.

robotblockset.cameras.image_transform.HWCImageType

an image with shape (H,W,C)

class robotblockset.cameras.image_transform.ImageTransform(input_shape: Tuple[int, int, int] | Tuple[int, int])[source]

Bases: ABC

property shape: Tuple[int, int, int] | Tuple[int, int]

The shape of the transformed image.

Returns:

ImageShapeType

Return type:

The shape of the transformed image.

transform_image(image: ndarray) → ndarray[source]

Apply the image transform to an image to get a new image.

Parameters:

image (HWCImageType) – The original image, it will be unaffected by the transform.

Raises:

NotImplementedError – Subclasses must implement this method.:

Returns:

HWCImageType

Return type:

The new, transformed image.

transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a point into transformed-image coordinates.

Transform the coordinates of a point from original image to transformed image.

reverse_transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a transformed-image point back to the source.

Transform the coordinates of a point in the transformed image back to the original image.

class robotblockset.cameras.image_transform.ComposedTransform(transforms: List[ImageTransform])[source]

Bases: ImageTransform

property shape: Tuple[int, int, int] | Tuple[int, int]

The shape of the transformed image.

Returns:

ImageShapeType

Return type:

The shape of the transformed image.

transform_image(image: ndarray) → ndarray[source]

Apply the image transform to an image to get a new image.

Parameters:

image (HWCImageType) – The original image, it will be unaffected by the transform.

Raises:

NotImplementedError – Subclasses must implement this method.:

Returns:

HWCImageType

Return type:

The new, transformed image.

transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a point into transformed-image coordinates.

Transform the coordinates of a point from original image to transformed image.

reverse_transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a transformed-image point back to the source.

Transform the coordinates of a point in the transformed image back to the original image.

robotblockset.cameras.image_transform.crop(image: ndarray, x: int, y: int, w: int, h: int) → ndarray[source]

Crop a rectangular region from an image.

Parameters:

  • image (HWCImageType) – Image to crop.

  • x (int) – X-coordinate of the top-left crop corner.

  • y (int) – Y-coordinate of the top-left crop corner.

  • w (int) – Crop width in pixels.

  • h (int) – Crop height in pixels.

class robotblockset.cameras.image_transform.Crop(input_shape: Tuple[int, int, int] | Tuple[int, int], x: int, y: int, w: int, h: int)[source]

Bases: ImageTransform

property shape: Tuple[int, int, int] | Tuple[int, int]

The shape of the transformed image.

Returns:

ImageShapeType

Return type:

The shape of the transformed image.

transform_image(image: ndarray) → ndarray[source]

Apply the image transform to an image to get a new image.

Parameters:

image (HWCImageType) – The original image, it will be unaffected by the transform.

Raises:

NotImplementedError – Subclasses must implement this method.:

Returns:

HWCImageType

Return type:

The new, transformed image.

transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a point into transformed-image coordinates.

Transform the coordinates of a point from original image to transformed image.

reverse_transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a transformed-image point back to the source.

Transform the coordinates of a point in the transformed image back to the original image.

class robotblockset.cameras.image_transform.Resize(input_shape: Tuple[int, int, int] | Tuple[int, int], h: int, w: int, round_transformed_points: bool = True)[source]

Bases: ImageTransform

Create a new Resize transform.

Note: Transforming a point to or from a resized image can lead to non-integer coordinates. Pixel coordinates

are however often expected to be integers, e.g. by the OpenCV draw functions. So by default, this class will round transformed points to the nearest integer. If you want to avoid the errors introduced by rounding, you can set round_transformed_points to False to get the exact transformed points as floats.

Parameters:

  • input_shape (ImageShapeType) – Shape of the images that will be resized.

  • h (int) – Height of the resized image.

  • w (int) – Width of the resized image.

  • round_transformed_points (bool, optional) – Whether to round transformed points to the nearest integer.

__init__(input_shape: Tuple[int, int, int] | Tuple[int, int], h: int, w: int, round_transformed_points: bool = True)[source]

Create a new Resize transform.

Note: Transforming a point to or from a resized image can lead to non-integer coordinates. Pixel coordinates

are however often expected to be integers, e.g. by the OpenCV draw functions. So by default, this class will round transformed points to the nearest integer. If you want to avoid the errors introduced by rounding, you can set round_transformed_points to False to get the exact transformed points as floats.

Parameters:

  • input_shape (ImageShapeType) – Shape of the images that will be resized.

  • h (int) – Height of the resized image.

  • w (int) – Width of the resized image.

  • round_transformed_points (bool, optional) – Whether to round transformed points to the nearest integer.

property shape: Tuple[int, int, int] | Tuple[int, int]

The shape of the transformed image.

Returns:

ImageShapeType

Return type:

The shape of the transformed image.

transform_image(image: ndarray) → ndarray[source]

Apply the image transform to an image to get a new image.

Parameters:

image (HWCImageType) – The original image, it will be unaffected by the transform.

Raises:

NotImplementedError – Subclasses must implement this method.:

Returns:

HWCImageType

Return type:

The new, transformed image.

transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a point into transformed-image coordinates.

Transform the coordinates of a point from original image to transformed image.

reverse_transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a transformed-image point back to the source.

Transform the coordinates of a point in the transformed image back to the original image.

class robotblockset.cameras.image_transform.Rotate90(input_shape: Tuple[int, int, int] | Tuple[int, int], num_rotations: int = 1)[source]

Bases: ImageTransform

Rotate an image by multiples of 90 degrees.

Create a new Rotate transform.

Parameters:

num_rotations (int, optional) – the number of 90-degree rotations to apply. Positive values rotate counter-clockwise.

__init__(input_shape: Tuple[int, int, int] | Tuple[int, int], num_rotations: int = 1)[source]

Create a new Rotate transform.

Parameters:

num_rotations (int, optional) – the number of 90-degree rotations to apply. Positive values rotate counter-clockwise.

property shape: Tuple[int, int, int] | Tuple[int, int]

The shape of the transformed image.

Returns:

ImageShapeType

Return type:

The shape of the transformed image.

transform_image(image: ndarray) → ndarray[source]

Apply the image transform to an image to get a new image.

Parameters:

image (HWCImageType) – The original image, it will be unaffected by the transform.

Raises:

NotImplementedError – Subclasses must implement this method.:

Returns:

HWCImageType

Return type:

The new, transformed image.

transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a point into transformed-image coordinates.

Transform the coordinates of a point from original image to transformed image.

reverse_transform_point(point: Tuple[int, int] | Tuple[float, float]) → Tuple[int, int] | Tuple[float, float][source]

Map a transformed-image point back to the source.

Transform the coordinates of a point in the transformed image back to the original image.

Attributes

HWCImageType

an image with shape (H,W,C)

Functions

crop(image, x, y, w, h)

Crop a rectangular region from an image.

Classes

ComposedTransform(transforms)

Crop(input_shape, x, y, w, h)

ImageTransform(input_shape)

Resize(input_shape, h, w[, ...])

Create a new Resize transform.

Rotate90(input_shape[, num_rotations])

Rotate an image by multiples of 90 degrees.