Inbuilt Video analytics module in AWS Greengrass

Kindly suggest. It's a bit urgent

Hi,
To answer your questions:

1. Greengrass have a list of connectors at https://docs.aws.amazon.com/greengrass/latest/developerguide/connectors-list.html . There is an MLImageCassification connector that you might be interested in, however it requires that the ML model is trained by the SageMaker image classification algorithm . As for " capture live video stream from an IP surveillance camera ", Greengrass doesn't have a connector for this. However you can experiment to create a pinned lambda for this function.
   2 & 3. I don't have the full context of your project so my answer might be inaccurate. From my understanding this should be achievable. You can write your lambda to include the customized object detection model , and have the lambda take camera input every second. The lambda can publish output to specific topics and Greengrass can configure the topic route.

Hi Shirley,
Thanks for your valuable inputs. By pinned lambda, do you mean to write a custom lambda function that will be able to read the streaming video? Also, is it possible to push an existing docker image to AWS Container repository and pull the same into a lambda deployed in AWS Greengrass?

Edited by: Am-400 on Sep 6, 2020 10:30 PM

Hi,

Pinned lambda is a long-lived lambda, detailed explanation is here https://docs.aws.amazon.com/greengrass/latest/developerguide/lambda-functions.html#lambda-lifecycle . You probably need to configure that lambda's containerization, so that it has access to camera/IO/etc. https://docs.aws.amazon.com/greengrass/latest/developerguide/lambda-group-config.html#lambda-containerization-considerations

You can pull docker image into your Greengrass device using the docker connector https://docs.aws.amazon.com/greengrass/latest/developerguide/docker-app-connector.html.

Thanks