Using the solution

Local inference with images from a camera

This section is a guidance to use the VIAI Edge solution running the model locally. It is a foundation, which should be modified as needed for each particular use case.

To acquire live images from the camera and feed them to the ML container to obtain an inference, run the following steps:

Run on Edge Server

  1. If not done earlier, export the ‘NAMESPACE` where the VIAI solution is deployed.

     export NAMESPACE=<your_namespace>

    The default value is viai-edge

  2. Open a shell in the camera integration container.

     kubectl exec -it viai-camera-integration-0 -n $NAMESPACE -- /bin/bash

  3. Execute the camera app to request a single image from the camera, and pass that to the ML model container.

    You will need the service IP or service name and port where the ML model is deployed from the previous steps.

    This example is for a Genicam camera:

     export ML_HOST=<your service name>
 export ML_PORT=8602

 python3 camera_client.py --protocol genicam --gentl /var/lib/viai/camera-config/FLIR_GenTL_Ubuntu_20_04_x86_64.cti \
     --cfg_write --cfg_write_file /var/lib/viai/camera-config/flir-ax5-recommended.cfg \
     --device_id ax5 --mode single --ml --ml_host ${ML_HOST} \
     --ml_port ${ML_PORT}

    Where:

    • ML_HOST is the name of your service name.
    • ML_PORT is the port of your service, 8602 by default.

    This example is for a USB camera:

     export ML_HOST=<your service name>
 export ML_PORT=8602
 export ADDRESS=<your camera address>
 export DEVICE_ID=<your device id>

 python3 camera_client.py --protocol usb --address ${ADDRESS} \
     --device_id ${DEVICE_ID} --mode single --ml --ml_host ${ML_HOST} --ml_port ${ML_PORT}

    Where:

    • ML_HOST is the name of your service name.
    • ML_PORT is the port of your service, 8602 by default.
    • ADDRESS is the address of the camera, usually /dev/video0. Check this page to find your camera address.
    • DEVICE_ID is the device ID of the camera, usually cam1. Check this page to find your camera device_id.

    The app should output the ML model’s inference results, which should look similar to this:

     {'predictionResult': {'annotationsGroups': [{'annotationSet': {'name': 'projects/199851883686/locations/us-central1/datasets/106648229847760896/annotationSets/5392255711264636928', 'displayName': 'Predicted Classification Labels', 'classificationLabel': {}, 'createTime': '2022-02-15T12:04:13.827789Z', 'updateTime': '2022-02-15T12:04:13.902274Z'},
 'annotations': [{'name': 'localAnnotations/0', 'annotationSpecId': '2438893262023426048', 'annotationSetId': '5392255711264636928', 'classificationLabel': {'confidenceScore': 0.52206558}, 'source':
 {'type': 'MACHINE_PRODUCED', 'sourceModel': 'projects/199851883686/locations/us-central1/solutions/7580859994631831552/modules/6066366290254102528/models/4464967186816958464'}, 'annotationSpecDisplayName': 'defect'}]}]}, 'predictionLatency': '0.035581135s'}

    In the output you can see the result as annotationSpecDisplayName (which is ‘defect’ in this example) and the confidenceScore and predictionLatency among others.

    Note: The response ‘annotiationSpecDisplayName’ from the VIAI model will always be defect, it is the ‘confidenceScore’ that determines how confident the model is about the image processed showing a defect (the closer to ‘1’, the more confidence on a defect present).