Skip To Content
ArcGIS Developer

Export Training Data For Deep Learning


Export Training Data For Deep Learning

The ExportTrainingDataforDeepLearning operation is designed to generate training sample image chips from the input imagery data with labeled vector data or classified images. The output of this service tool is the data store string where the output image chips, labels, and metadata files are going to be stored.

Request parameters



The image that will be classified. This can be specified as the portal item ID, image service URL, cloud raster dataset, shared raster dataset, a feature service with image attachments, or a raster dataset or image collection in the data store. At least one type of input must be provided in the JSON object. If multiple inputs are given, the itemId takes priority.

Syntax: JSON object describes the inputRaster.


//Portal Item ID
inputRaster={"itemId": <portal item id>}

//Image Service URL
inputRaster={"url": <image service url>}

//Service Properties
inputRaster={"serviceProperties":{"name":"testrasteranalysis","serviceUrl":"https://<server name>/server/rest/services/Hosted/testrasteranalysis/ImageServer"},"itemProperties":{"itemId":"8cfbd3ec25584d0d8fed23b8ff7c43b","folderId":"sdfwerfbd3ec25584d0d8f4"}}


This is the output location for training sample data. This can be specified as the output folder name, a file share raster data store path, a file share data store path, or a shared file system path.


//Output folder name

//File share data store path

//File share raster data store path

//File share path


The labeled data, either in a feature service or an image service. Vector inputs should follow a training sample format as generated by the ArcGIS Pro Training Sample Manager, whereas raster inputs should follow a classified raster format as generated by the Classify Raster tool.

Syntax: JSON object describes the inputClassData.


//Portal Item ID
{"itemId": <portal item id>}

//Service URL
{"url": <image or feature service url>}

//Service Properties
{"serviceProperties":{"name":"testrasteranalysis","serviceUrl":"https://<server name>/server/rest/services/Hosted/testrasteranalysis/ImageServer"},"itemProperties":{"itemId":"8cfbd3ec25584d0d8fed23b8ff7c43b", "folderId":"sdfwerfbd3ec25584d0d8f4"}}

Specifies the raster format that will be used for the image chip outputs.

Values: TIFF | PNG | JPEG | MRF (Meta Raster Format)



The size of the image chips. This is specified as a name value pair for x and y dimension values.

Syntax: A JSON object describes the tileSize.



The distance to move in the x and y when creating the next image chip. This is specified as a name value pair for x and y dimension values. When stride is equal to the tile size, there will be no overlap. When stride is equal to half of the tile size, there will be 50 percent overlap.

Syntax: A JSON object describes the strideSize.



Specifies the format of the output metadata labels.

If your input training sample data is a feature class layer, such as a building layer or standard classification training sample file, use the KITTI or PASCAL VOC rectangles option. The output metadata is a .txt file or .xml file containing the training sample data contained in the minimum bounding rectangle. The name of the metadata file matches the input source image name. If your input training sample data is a class map, use the Classified Tiles option as your output metadata format.


  • KITTI_rectangles—The metadata follows the same format as the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) Object Detection Evaluation dataset. The KITTI dataset is a vision benchmark suite. The label files are plain text files. All values, both numerical and strings, are separated by spaces, and each row corresponds to one object. For more information, see KITTI metadata format. This format is used for object detection.
  • PASCAL_VOC_rectangles (Default)—The metadata follows the same format as the Pattern Analysis, Statistical Modeling and Computational Learning, Visual Object Classes (PASCAL_VOC) dataset. The PASCAL VOC dataset is a standardized image dataset for object class recognition. The label files are XML files and contain information about image name, class value, and bounding boxes. For more information, see PASCAL Visual Object Classes. This format is used for object detection. This is the default.
  • Classified_Tiles—This option will output one classified image chip per input image chip. No other metadata for each image chip is used. Only the statistics output has more information on the classes, such as class names, class values, and output statistics. This format is used for pixel classification.
  • RCNN_Masks—This option will output image chips that have a mask on the areas where the sample exists. The model generates bounding boxes and segmentation masks for each instance of an object in the image. It's based on Feature Pyramid Network (FPN) and a ResNet101 backbone in the deep learning framework model. This format is used for object detection.
  • Labeled_Tiles—Each output tile will be labeled with a specific class. This format is used for object classification.
  • MultiLabeled_Tiles—Each output tile will be labeled with one of more classes. For example, a residence may be labeled as containing a pool and also solar panels. This format is used for object classification.
  • Export_Tiles—The output will be image chips with no label. This format is used for image enhancement techniques such as super resolution.

PASCAL_VOC_rectangles example

<?xml version=”1.0”?>
- <layout>
    - <part>
       - <bndbox>

The field that contains the class values. If no field is specified, the system searches for a value or classvalue field. If the feature does not contain a class field, the system determines that all records belong to one class.



The radius for a buffer around each training sample to delineate a training sample area. This allows you to create circular polygon training samples from points.



A polygon feature class that delineates the area where image chips will be created. Only image chips that fall completely within the polygons will be created.


inputMaskPolygons={"itemId": <portal item id>}
inputMaskPolygons={"url": <feature service url>}

The rotation angle that will be used to generate additional image chips. An image chip will be generated with a rotation angle of 0, which means no rotation. It will then be rotated at the specified angle to create an additional image chip. The same training samples will be captured at multiple angles in multiple image chips for data augmentation. The default rotation angle is 0.



Specifies the type of reference system to be used to export the image tiles, either MAP_SPACE or PIXEL_SPACE. Choose MAP_SPACE when the input image is in the map_based coordinate system. This is the default value. PIXEL_SPACE should be used when the input image is in image space, with no rotation and no distortion.



Specifies how raster items in an image service will be processed. When false, all raster items in the image service will be mosaicked together and processed. This is the default option. When true, all raster items in the image service will be processed as a separate image.

Values: true | false


Specifies whether to blacken the pixels around each object or feature in each image tile. This parameter applies only when the metadata format is set to Labeled_Tiles and an input feature class or classified raster has been specified. When false, pixels surrounding objects or features will not be blackened. This is the default. When true, pixels surrounding objects or features will be blackened.

Values: true | false


Specifies whether to crop the exported tiles such that they are all the same size. This parameter applies only when the metadata format is set to Labeled_Tiles and an input feature class or classified raster has been specified. When true, exported tiles will be the same size and will center on the feature. This is the default. When false, exported tiles will be cropped such that the bounding geometry surrounds only the feature in the image tile.

Values: true | false


Contains settings that affect task execution. This task has the following settings:

  • extent—A bounding box that defines the analysis area.
  • cellSize—The output raster will have the resolution specified by cell size.
  • exportAllTiles—Choose if the training sample image chips with overlapped label data will be exported. If true, all image chips, including those that do not overlap labeled data, will be exported. This is the default. If false, only the image chips that overlap the labeled data will be exported.
  • startIndex—Allows you to set the start index for the sequence of image chips. This appends more image chips to an existing sequence. The default value is 0.

The response format. The default response format is html.

Values: html | json | pjson

Additional KITTI metadata format information

The table below describes the 15 values in the KITTI metadata format. Only 5 of the possible 15 values are used in the tool: the class name (in column 1) and the minimum bounding rectangle composed of four image coordinate locations (columns 5–8). The minimum bounding rectangle encompasses the training chip used in the deep learning classifier.



Class value

The class value of the object listed in the stats.txt file.





The two-dimensional bounding box of objects in the image, based on a 0-based image space coordinate index. The bounding box contains the four coordinates for the left, top, right, and bottom pixels.



Example usage

Below is a sample GET request URL for ExportTrainingDataforDeepLearning:{"itemId":89964029c5354407a4f817187144be42}&outputLocation=/rasterStores/myrasterstore/rooftoptrainingsamples&inputClassData={"itemId":66b1f5fa24b14217a1129f8ab688386a}&chipFormat=TIFFtileSize={"x":256,"y":256}&strideSize={"x":128,"y":128}&metadataFormat=KITTI_rectangles&classValueField=&bufferRadius=1&inputMaskPolygons=&rotationAngle=0&referenceSystem=MAP_SPACE&processAllRasterItems=false&blackenAroundFeature=false&fixChipSize=true&f=pjson

Below is a sample POST request URL for ExportTrainingDataforDeepLearning

POST /webadaptor/rest/services/System/RasterAnalysisTools/GPServer/ExportTrainingforDeepLearning HTTP/1.1
Content-Type: application/x-www-form-urlencoded
Content-Length: []



When you submit a request, the task assigns a unique job ID for the transaction.


"jobId": "<unique job identifier>",
"jobStatus": "<job status>"

After the initial request is submitted, you can use the jobId to periodically check the status of the job and messages as described in Checking job status. Once the job has successfully completed, you use the jobId to retrieve the results. To track the status, you can make a request of the following form:

https://<raster analysis tools url>/ExportTrainingDataforDeepLearning/jobs/<jobId>

When the status of the job request is esriJobSucceeded, you can access the results of the analysis by making a request of the following form:

https://<raster analysis tools url>/ExportTrainingDataforDeepLearning/jobs/<jobId>/results/outLocation

JSON Response example

The response returns the outLocation parameter, which has properties for parameter name, data type, and value. The content of the value is always the output data store item's itemId or URL. The parameter provides the output location of the training data.

  "paramName": "outLocation",
  "dataType": "GPString",
  "value": {
    "uri": "/rasterStores/myrasterstore/rooftops"