implementing Amazon Rekognition on AWS Greengrass for Edge computing

I am working on implementing Rekognition on AWS Greengrass for testing it for edge computing. My main goal is detecting objects by sending from a RaspberryPi and measuring the latency on edge computing using Greengrass. I setup the IoT core, components and ...but don't know how should I combine Rekognition with the GG to compute locally at the edge. I'd be grateful if you share any experience on this regard with me.

P.S: I also completed the Lambda function and it is included in my GG group.