This article applies to these versions of LandingLens:

            LandingLens            LandingLens on Snowflake

LandingAI offers an optical character recognition (OCR) solution through LandingEdge and Docker. Use OCR to extract text from images to convert unstructured content into structured, actionable data. Our OCR solution supports English and Simplified Chinese character sets and can detect multiple languages in one image.

This OCR solution is available as an add-on for users on the LandingLens Enterprise plan.

When LandingEdge runs OCR on an image, the predicted text displays on the image and is included in the JSON response.

​
Activation Keys

Using OCR in LandingEdge or Docker requires an activation key. The same activation key can be used for both LandingEdge and Docker.

If you’ve purchased the OCR add-on and need your activation key, contact support@landing.ai.

​
OCR in LandingEdge

OCR is available in LandingEdge v2.9.1 and later. To use OCR in LandingEdge, first enter your activation key in the LandingEdge settings. Then, you can create an Inspection Point for OCR.

​
Enter the Activation Key in LandingEdge

Once you have an OCR activation key, follow the instructions below to add it to LandingEdge:

  1. Open LandingEdge.
  2. Click the Settings (gear) icon in the top right corner of the LandingEdge homepage.
  3. Click OCR.
  4. Enter the activation key in the OCR Activation Key field.
  5. Click Save.

​
Run OCR in LandingEdge

After you’ve entered the activation key in LandingEdge, you can set up an Inspection Point to run OCR. To do this, follow the instructions below:

  1. Create an Inspection Point in LandingEdge.
  2. Set up the Image Source and Inspection Start settings.
  3. Skip the Cloud Connection section.
  4. Select OCR from the Model drop-down menu.
  5. Set up the other settings if needed.
  6. Click Save Configuration.
  7. Click Run.
  8. When you run OCR in LandingEdge, the results display in the user interface. There is a box around each detected string of text and the predicted text displays above that box. The results are also included in the JSON response.

​
OCR with Docker

You can run the LandingAI OCR tool with the LandingAI Deploy Docker solution. OCR is available in LandingAI Docker v2.9.1 and later.

First, launch a Docker container with the LandingAI OCR model. To do this, include your LandingAI OCR activation key and the run-ocr command when launching a container. For example:

docker run -e LANDING_LICENSE_KEY="your_activation_key" -p 8000:8000 public.ecr.aws/landing-ai/deploy:latest run-ocr

After launching the Docker container with the OCR tool, you can run OCR on images the same way you would send images for inference. For detailed information, go to LandingAI Docker.

​
OCR JSON Response

The following table describes the objects in the JSON response for OCR models. An example response is here.

ElementDescription
typeThe name of the prediction type. This will always be OCR.
predictionsThis object contains the information for each string of text the model detected. Each prediction is a separate object nested in the predictions object.

Predictions are listed in alphanumeric order.
textThe predicted text of the string.
scoreThe confidence score for the prediction.
locationThis object contains the pair of x and y coordinates (in pixels) of each corner of the bounding box of the predicted string. Each coordinate is a separate object nested in the location object.

The coordinates start with the bottom left corner of the bounding box and go counterclockwise.
metadataThis element contains nested metadata from the image. If the Image Source is Folder Watcher, some data is populated by default. You can use scripts and web APIs to set or override the values.
image_idIf the Image Source is Folder Watcher, this is the file name of the image. Otherwise, this object is blank. You can use scripts and web APIs to set or override the values.
inspection_station_idThis element is blank by default. You can use scripts and web APIs to set or override the values.
location_idIf the Image Source is Folder Watcher, this is the directory that the image is in. Otherwise, this object is blank. You can use scripts and web APIs to set or override the values.
capture_timestampThe time and date that OCR was run on the image. You can use scripts and web APIs to set or override the values.
model_idThis will always be OCR.
latencyThe latency object includes the detailed timing statistics of the inference call.

Each key-value pair in the latency object represents a step in the inference process, and the duration of that step. All values are measured in seconds.

​
Example: OCR JSON Response

The following image and code snippet show the OCR model’s predictions. The model correctly predicted two strings of text: “ROAD” and “CLOSED”.

{
    "type": "OCR",
    "predictions": [
        {
            "text": "CLOSED",
            "score": 0.98919207,
            "location": [
                {
                    "x": 1639,
                    "y": 1712
                },
                {
                    "x": 1650,
                    "y": 1454
                },
                {
                    "x": 2687,
                    "y": 1496
                },
                {
                    "x": 2676,
                    "y": 1755
                }
            ]
        },
        {
            "text": "ROAD",
            "score": 0.9103826,
            "location": [
                {
                    "x": 1778,
                    "y": 1368
                },
                {
                    "x": 1787,
                    "y": 1100
                },
                {
                    "x": 2542,
                    "y": 1126
                },
                {
                    "x": 2533,
                    "y": 1394
                }
            ]
        }
    ],
    "metadata": {
        "image_id": "malachi-brooks-SmgvToT3nbA-unsplash.jpg",
        "inspection_station_id": "",
        "location_id": "/Users/user/Desktop/folder",
        "capture_timestamp": "2024-08-14T18:27:28.286829-07:00"
    },
    "model_id": "OCR",
    "latency": {
        "decoding_s": 0.1738631,
        "preprocess_s": 0.0030414,
        "waiting_s": 0.0,
        "infer_s": 0.0826073,
        "postprocess_s": 0.0133133
    }
}