Available with Image Server
The Classify Pixels Using Deep Learning tool uses a deep learning model to classify the pixels in an imagery layer according to a defined list of labels indicating different classes.
The output is a hosted imagery layer.
Example
The Classify Pixels Using Deep Learning tool can be used as inputs for categorical change detection between different time periods. The classified thematic imagery layers produced as outputs can be used as the input imagery layers for measuring change over time. For example, the Classify Pixels Using Deep Learning tool can be used to create a thematic imagery layer for the suburbs of a major city for two time periods with the same classification theme. By comparing the thematic imagery layers produced, the transition of areas between labeled classes can be measured and quantified.
Usage notes
Classify Pixels Using Deep Learning tool includes configurations for input layer, model settings, and result layer.
Input layers
The Input layers group includes the following parameters:
- Input layer is the imagery layer or layer that will be used for the classification. The imagery layer selected should be based on the requirements for the deep learning model that will be used to classify the pixels.
- Processing mode describes how the raster items in the imagery layer will be processed all together or individually. Processing mode includes the following options:
- Process as mosaicked image—All raster items in the mosaic dataset or image service will be mosaicked together and processed. This is the default.
- Process all raster items separately—All raster items in the mosaic dataset or image service will be processed as separate images.
Model settings
The Model settings group includes the following parameters:
- Model for pixel classification indicates which deep learning model to be used to classify the pixels. The deep learning model needs to be located on ArcGIS Online to be selected in the tool. You can select your own model, publicly available in ArcGIS Online, or from ArcGIS Living Atlas of the World.
- Model arguments specifies the function arguments defined in the Python raster function class. Additional deep learning parameters and arguments for experiments and refinement are listed, such as a confidence threshold for adjusting the sensitivity. The names of the arguments are populated from the Python module.
Result layer
The Result layer group includes the following parameters:
- Output name determines the name of the layer that is created and added to the map. The name must be unique. If a layer with the same name already exists in your organization, the tool will fail and you will be prompted to use a different name.
- Save in folder specifies the name of a folder in My content where the result will be saved.
Environments
Analysis environment settings are additional parameters that affect a tool's results. You can access the tool's analysis environment settings from the Environment settings parameter group.
This tool honors the following analysis environments:
- Output coordinate system
- Processing extent
Note:
The default processing extent in Map Viewer is Full extent. This default is different from Map Viewer Classic in which Use current map extent is enabled by default.
- Snap raster
- Cell size
- Parallel processing factor
- Processor type
Outputs
The output is a classified thematic imagery layer based on the classification scheme defined in the deep learning model.
Usage requirements
This tool requires the following licensing and configurations:
- Creator or GIS Professional user type
- Publisher or Administrator role, or an equivalent custom role
- ArcGIS Image Server configured for deep learning raster analytics
Resources
Use the following resources to learn more:
- Classify Pixels Using Deep Learning in ArcGIS REST API
- classify_pixels function in ArcGIS API for Python
- Classify Pixels Using Deep Learning in ArcGIS Pro