How to use Object format for images

# component.add_input('vision_payload', 'String')
# Use this instead:
component.add_input('vision_payload', 'Object')

In entrypoint.py in process_input:

# vision_payload = json.loads(data['vision_payload'])
# pil_image = payload.get_image_from_vision_mqtt_payload(vision_payload)

# Use this instead:
pil_image = payload.get_image_from_vision_zmq_dict(data["vision_payload"])

If you want to return the thumbnail, add the appropriate output.

component.add_input('vision_payload', 'Object')
component.add_output('prediction', 'String')
# Add this line:
component.add_output('thumbnail', 'Object')

In entrypoint.py in process_input:

return {
    "prediction": str(prediction),
    "ic_probability": metric_output(probability),
    # Add this line:
    "thumbnail": payload.create_image_output(pil_image),
}

AI Inference Server version 1.6 imposes restrictions on the dictionaries returned by the inference wrappers for output variables of type Object. The returned dictionary must contain a metadata string and a binary sequence. The metadata and the binary sequence can have any key in the dict, but must be of type str and bytes respectively.For details, refer to the Returning results howto guide.

⚠️ Note The structure of dictionaries that is received as pipeline input is different from the dictionary structure that is required as component output. For details, see the section Processing Images.