Unable to classify pixels in an imagery layer in ArcGIS Enterprise and Online Map Viewer

Description

When using the Classify Pixels Using Deep Learning tool in ArcGIS Enterprise or Online Map Viewer, the imagery layer does not appear as an available input. In some instances, using an imagery layer from ArcGIS Living Atlas of the World as the input results in execution failure, as shown in the image below.

Cause

The imagery layer is either published as a map image layer or not hosted by the user, both of which limit pixel-level access. Deep learning workflows require dynamic imagery layers that provide direct access to pixel values and metadata to perform pixel-based analysis, object detection, image classification, and semantic segmentation.

Solution or Workaround

Publish as an image service

Publish the imagery dataset as an image service instead of a map service to ensure compatibility with deep learning tools. Refer to ArcGIS Pro: Publish an image service for instructions.

Use hosted imagery as the input imagery layer

Ensure the input imagery layer is hosted by the user. Esri recommends not to use layers from shared sources, such as ArcGIS Living Atlas of the World, unless they explicitly support deep learning workflows and provide the necessary analysis capabilities.

Perform deep learning workflows in ArcGIS Pro

Classify pixels in the imagery layer in ArcGIS Pro. Refer to How To: Classify multispectral imagery pixels using Pix2Pix in ArcGIS Pro for instructions. The output raster dataset can then be published as an image service to the ArcGIS Enterprise portal or ArcGIS Online.