Configure Image Transformations
Transformation blocks are typically used at the beginning of a pipeline to transform data before sending it to a model.
Image transformation blocks create permanent changes to a working copy of an image in the pipeline.
Image transformations such as cropping allow the user to make changes to data as it moves through the pipeline.
Crop the image to the specified height and width. Center cropping will make center based adjustments, while fixed cropping will cut from a specific location on the image defined by you. Cropping will permanently modify a working copy of the image which will be returned in the pipeline outputs. The original source image will not be affected.
Click to configure an image crop processing block to add to your pipeline.
To add image cropping to your pipeline:
- 1.Navigate to the "Pipeline Blocks" tab.
- 2.Under the Image Transformation section, click on the Image Crop option to configure this processing block.
- 3.Select either the Center Crop or Fixed Crop option.
Set a maximum height and width to center crop your images.
An example of a center-crop transformation on an image.
Set the maximum height and width of the image crop and the position of its top left corner using X and Y coordinates. Cropping will be relative to the top left corner position specified.
An example of a fixed-crop transformation transformation on an image.
This advanced setting is available for both center and fixed crop options. Minimum Visibility defines the percentage of a detection that must be within the crop to be retained. Ex. If minimum visibilty is set at 50% and the crop action removes 90% of that instance in an image, we will no longer include that instance as a detection. By default, this value is set to 20%
When finished configuring your image crop settings, click "Add to Pipeline" to add this processing block.
The Image Resize block can resize images to a fixed size, which will change the aspect ratio of the image, or it can adjust the longest side of the image, which will preserve the aspect ratio. Predictions made in the pipeline before this block will be scaled along with the image.
Use care when resizing images before model blocks, since making images smaller than the model’s preferred size can decrease prediction accuracy.
Click to configure an image resize processing block to add to your pipeline.
To add image resize to your pipeline:
- 1.Navigate to the "Pipeline Blocks" tab.
- 2.Under the Image Transformation section, click on the Image Resize option to configure this processing block.
- 3.Select either Fixed Resize or Longest Size.
Fixed Resize will scale each image up or down to the dimensions you set, regardless of the image’s starting aspect ratio. Set the height and width values to the desired dimensions. For best performance, we recommend values between 200px through 800px.
An example of a fixed resize transformation on an image.
Longest Side resizing will scale each image’s longest edge up or down to the dimension you set, while preserving the image’s starting aspect ratio. Set the Size value to scale the longest side to this size. For best performance, we recommend values between 200px through 800px.
An example of a longest side resize transformation on an image.
When finished configuring your image resize settings, click "Add to Pipeline" to add this processing block.
Last modified 7mo ago