Performance Issue on Edge

[sakh251]

Hello,
We deployed lookout for vision trained model on edge device using Greengrass. Even with multi-threading and multi-processing, we can not utilize the GPU. It is always less than 5% 7%. I tried with single shared channel between all thread or separate channels. As we checked it seems the inference is performed sequentially on gRPC server. Since we have several cameras we need to have better performance.

Thank you
IoT Greengrass Core software version: 2