Defining JSON lines for image classification

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 classification model, the JSON line must include an image classification that is either normal or an anomaly. A JSON line is in SageMaker AI Ground Truth Classification Job Output format. A manifest file is made of one or more JSON lines, one for each image that you want to import.

To create a manifest file for classified images

1. Create an empty text file.

2. Add a JSON line for each image the that you want to import. Each JSON line should look similar to the following:

   {"source-ref":"s3://lookoutvision-console-us-east-1-nnnnnnnnnn/gt-job/manifest/IMG_1133.png","anomaly-label":1,"anomaly-label-metadata":{"confidence":0.95,"job-name":"labeling-job/testclconsolebucket","class-name":"normal","human-annotated":"yes","creation-date":"2020-04-15T20:17:23.433061","type":"groundtruth/image-classification"}}
   

3. Save the file.

   Note

   You can use the extension .manifest, but it is not required.

4. Create a dataset using the manifest file that you created. For more information, see Creating a manifest file.

Classification JSON lines

In this section, you learn how to create a JSON line that classifies an image as normal or anomalous.

Anomaly JSON line

The following JSON line shows an image that is labeled as an anomaly. Note that the value of class-name is anomaly.

{
    "source-ref": "s3: //bucket/image/anomaly/abnormal-1.jpg",
    "anomaly-label-metadata": {
        "confidence": 1,
        "job-name": "labeling-job/auto-label",
        "class-name": "anomaly",
        "human-annotated": "yes",
        "creation-date": "2020-11-10T03:37:09.600",
        "type": "groundtruth/image-classification"
    },
    "anomaly-label": 1
}                
                

Normal JSON line

The following JSON line shows an image labeled as normal. Note that the value of class-name is normal.

{
    "source-ref": "s3: //bucket/image/normal/2020-10-20_12-14-55_613.jpeg",
    "anomaly-label-metadata": {
        "confidence": 1,
        "job-name": "labeling-job/auto-label",
        "class-name": "normal",
        "human-annotated": "yes",
        "creation-date": "2020-11-10T03:37:09.603",
        "type": "groundtruth/image-classification"
    },
    "anomaly-label": 0
}

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 (0 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. The field name must be the same as the label attribute with -metadata appended.

confidence

(Optional) Currently not used by Amazon Lookout for Vision. If you do specify a value, use a value of 1.

job-name

(Optional) A name that you choose for the job that processes the image.

class-name

(Required) If the image contains normal content, specify normal, otherwise specify anomaly. If the value of class-name is any other value, the image is added to the dataset as an unlabeled image. To label an image, see Adding images to your dataset.

human-annotated

(Required) Specify "yes", if the annotation was completed by a human. Otherwise, specify "no".

creation-date

(Optional) The Coordinated Universal Time (UTC) date and time that the label was created.

type

(Required) The type of processing that should be applied to the image. For image-level anomaly labels, the value is "groundtruth/image-classification".