Skip to main content Link Search Menu Expand Document (external link)
Hidale.com Docs

@mlcrop attribute

:electric_plug: dp.kinect3 dp.oak
:information_source: signature
mlcrop { off | center | centerpad | XMIN YMIN XMAX YMAX }
:house: values
mlcrop off default
:bulb: examples
@mlcrop centerpad
@mlcrop 160 140 479 379

Crop the machine learning input tensor. This overrides the default behavior to automatically scale the input to the model’s tensor input size requirements without preserving aspect ratio.

  • off resize to input requirement, may change aspect ratio
  • center resize and center crop to input requirement, loose cropped areas, keep aspect ratio
  • centerpad resize and pad the color image to input requirement, keep aspect ratio
  • XMIN YMIN XMAX YMAX four integer crop region, resize to input requirement, may change aspect ratio
    e.g. @mlcrop 160 140 479 379 crops tensor to have inclusive columns 160->479 and rows 140->379, then resized by @mlinput requirement.

Visit @mlinput for a diagram of the crop pipeline.

@mlcrop can be changed anytime. It can automated by sending messages like mlcrop 12 34 56 78 to our plugins. This enables dynamic focus, zoom, and crop of data based on real-time algorithms and choices. For example, if a model identifies a human in an image and provides a bounding box, then the bounding box pixel coordinates can be used to focus a model on only that human by constructing a mlcrop message and sending it to the plugin.