Use the model

You can use this model in the Classify Pixels Using Deep Learning tool available in the Image Analyst toolbox in ArcGIS Pro. Follow the steps below to use the model for classifying land cover in images.

Recommended imagery configuration

The recommended imagery configuration is as follows:

• Imagery—Landsat 8 Collection 2 Level-2 imagery
• Resolution—30 meters
• Supported configurations are as follows:
  • Landsat 8 Collection 2 Level-2 imagery—Surface Reflectance in form of a raster, mosaic dataset, or image service
• When using a raster product, ensure that you choose the Surface Reflectance product while adding the imagery to your map. When using a mosaic dataset, ensure that you choose the Landsat 8 raster type and Surface Reflectance processing template when adding rasters.

Classify land cover

Use the following steps to classify land cover from the imagery:

Data Preparation:
1. Prepare data according to the following product type:
   • Raster Product
   1. Download the Land Cover Classification (Landsat 8) model to your local machine.
   2. Navigate to the folder housing the Landsat 8 Collection 2 Level-2 imagery data. Expand the folder and locate the raster product.
   3. Expand the raster product provided as an MTL.txt file and select the Surface Reflectance derived raster dataset.

      

   • Mosaic Dataset
   1. Create or locate a Mosaic Dataset within a File Geodatabase. Right-click the Mosaic Dataset and select Add Rasters to Mosaic Dataset.
   2. In the tool dialogue, select the Landsat 8 raster type and the Surface Reflectance processing template and run the tool to completion.

      

   • Raster Dataset
   1. Traditionally Landsat 8 Collection 2 data is provided as a series of .tif dataset with a .txt metadata file required for use as a raster product. If this metadata distribution is not available, you can create a multiband image through either the Composite Bands geoprocessing tool or Composite Bands raster function. This can be used as input for inferencing.

      

      

Data Processing:
2. Zoom to an area of interest.

   

3. Browse to Tools on the Analysis tab.

   

4. Click the Toolboxes tab in the Geoprocessing pane, select Image Analyst Tools, and browse to the Classify Pixels Using Deep Learning tool under Deep Learning.

   

5. Set the variables on the Parameters tab as follows:
   1. Input Raster—Select the imagery as discussed above.
   2. Output Classified Raster—Set the output raster dataset that will contain the classification results.
   3. Model Definition—Select the pretrained or fine-tuned model .dlpk file. For this use case utilize the Land Cover Classification (Landsat 8) model downloaded previously.
   4. Processing Mode—Select the Process as mosaicked image mode.
   5. Arguments (optional)—Change the values of the arguments if required.
      • padding—Number of pixels at the border of image tiles from which predictions are blended for adjacent tiles. Increase its value to smooth the output while reducing edge artifacts. The maximum value of the padding can be half of the tile size value.
      • batch_size—Number of image tiles processed in each step of the model inference. This depends on the memory of your graphic card.
      • predict_background—If set to True, background class is also classified.
      • test_time_augmentation—Performs test time augmentation while predicting. If true, predictions of flipped and rotated variants of the input image will be merged into the final output.
      • tile_size—The width and height of image tiles into which the imagery is split for prediction.
      • landsat_imagery_level—The processing level of input Landsat 8 imagery. The default value is 2 for Landsat Collection 2 Level-2 imagery.

   

   Note:

   To access the model directly from ArcGIS Pro (supported in ArcGIS Pro 2.7 and later), click the browse button and search for the model.

   

6. Set the variables on the Environments tab as follows:
   1. Processing Extent—Select Current Display Extent or any other option from the drop-down menu.
   2. Cell Size (required)—Set the value to 30.

      The expected raster resolution is 30 meters.

   3. Processor Type—Select CPU or GPU.

      It is recommended that you select GPU, if available, and set GPU ID to specify the GPU to be used.

   

Visualization:
7. Click Run.

   The output layer is added to the map.