End of support notice: On October 31, 2025, AWS will discontinue support for Amazon Lookout for Vision. After October 31, 2025, you will no longer be able to access the Lookout for Vision console or Lookout for Vision resources. For more information, visit this blog post.

Defining JSON lines for image segmentation

You define a JSON line for each image that you want to use in an Amazon Lookout for Vision manifest file. If you want to create a segmentation model, The JSON line must include segmentation and classification information for the image. A manifest file is made of one or more JSON lines, one for each image that you want to import.

Segmentation JSON lines

In this section, you learn how to create a JSON line that includes segmentation and classification information for an image.

The following JSON line shows an image with segmentation and classification information. anomaly-label-metadata contains classification information. anomaly-mask-ref and anomaly-mask-ref-metadata contain segmentation information.

{
    "source-ref": "s3://path-to-image",
    "anomaly-label": 1,
    "anomaly-label-metadata": {
        "class-name": "anomaly",
        "creation-date": "2021-10-12T14:16:45.668",
        "human-annotated": "yes",
        "job-name": "labeling-job/classification-job",
        "type": "groundtruth/image-classification",
        "confidence": 1
    },
    "anomaly-mask-ref": "s3://path-to-image",
    "anomaly-mask-ref-metadata": {
        "internal-color-map": {
            "0": {
                "class-name": "BACKGROUND",
                "hex-color": "#ffffff",
                "confidence": 0.0
            },
            "1": {
                "class-name": "scratch",
                "hex-color": "#2ca02c",
                "confidence": 0.0
            },
            "2": {
                "class-name": "dent",
                "hex-color": "#1f77b4",
                "confidence": 0.0
            }
        },
        "type": "groundtruth/semantic-segmentation",
        "human-annotated": "yes",
        "creation-date": "2021-11-23T20:31:57.758889",
        "job-name": "labeling-job/segmentation-job"
    }
}               
                

JSON line keys and values

The following information describes the keys and values in an Amazon Lookout for Vision JSON line.

source-ref

(Required) The Amazon S3 location of the image. The format is "s3://BUCKET/OBJECT_PATH". Images in an imported dataset must be stored in the same Amazon S3 bucket.

anomaly-label

(Required) The label attribute. Use the key anomaly-label, or another key name that you choose. The key value (1 in the preceding example) is required by Amazon Lookout for Vision, but it isn't used. The output manifest created by Amazon Lookout for Vision converts the value to 1 for an anomalous image and a value of 0 for a normal image. The value of class-name determines if the image is normal or anomalous.

There must be corresponding metadata identified by the field name with -metadata appended. For example, "anomaly-label-metadata".

anomaly-label-metadata

(Required) Metadata about the label attribute. Contains classification information. The field name must be the same as the label attribute with -metadata appended.

anomaly-mask-ref

(Required) The Amazon S3 location of the mask image. Use anomaly-mask-ref for the key name or use a key name of your choosing. The key must end with -ref. The mask image must contain colored masks for each anomaly type internal-color-map. The format is "s3://BUCKET/OBJECT_PATH". Images in an imported dataset must be stored in the same Amazon S3 bucket. The mask image must be a Portable Network Graphic (PNG) format image.

anomaly-mask-ref-metadata

(Required) Segmentation metadata for the image. Use anomaly-mask-ref-metadata for the key name or use a key name of your choosing. The key name must end with -ref-metadata.