Expand description
Vision module.
Modules§
Macros§
- dispatch_
bool_ dtype - Dispatches connected components based on the bool dtype, binding a concrete
integer type to enable generic instantiations without extra trait bounds (after removing
ElementComparisonfromElement). - dispatch_
int_ dtype - Dispatches connected components based on the int dtype, binding a concrete
integer type to enable generic instantiations without extra trait bounds (after removing
ElementComparisonfromElement).
Structs§
- BoxBlur
- Blurs a batch of images with a uniform box filter, applied with a given
probability. Built from a
BoxBlurConfig. - BoxBlur
Config - Configuration for a
BoxBluraugmentation. - Connected
Stats - Stats collected by the connected components analysis
- Connected
Stats Options - Which stats should be enabled for
connected_components_with_stats. Currently only used by the GPU implementation to save on atomic operations for unneeded stats. - Connected
Stats Primitive - Primitive version of
ConnectedStats, to be returned by the backend. - Connected
Stats Primitive Metadata - Tensor metadata and ordinary fields for the corresponding backend extension value.
- Median
Blur - Blurs a batch of images with a square median filter, applied with a given
probability. Built from a
MedianBlurConfig. - Median
Blur Config - Configuration for a
MedianBluraugmentation. - Morph
Options - Options for morphology ops
- Morph
Options Builder - Use builder syntax to set the inputs and finish with
build(). - NmsOptions
- Non-Maximum Suppression options.
- Point
- 2D Point used for vision ops. Coordinates start at the top left.
- Size
- 2D size used for vision ops.
- Transform2D
- 2D point transformation
Enums§
- Border
Type - Morphology border type
- Connectivity
- Connected components connectivity
- Kernel
Shape - Shape of the structuring element
Traits§
- Bool
Vision Ops - Vision ops on bool tensors
- Color
Conversion - Color space conversion tensor operations.
- Connected
Components - Connected components tensor extensions
- Float
Vision Ops - Vision ops on float tensors
- IntVision
Ops - Vision ops on int tensors
- Morphology
- Morphology tensor operations
- Nms
- Non-maximum suppression tensor operations
- Vision
Backend - Vision capable backend, implemented by each backend
Functions§
- create_
structuring_ element - Create a structuring element tensor for use with morphology ops
- filter2d
- Depthwise convolution of a tensor with a 2D kernel.
- median
- Apply a depth-wise square median filter to the given images