Multi Camera Support for Panorama

0

Hello AWS,

Our company (aidant.ai) is considering using AWS Panorama as our appliance of choice to analyze video streams and provide AI inference.
We are currently using Nvidia RTX2070 GPUs at the local appliance and are able to run 4 video streams per GPU.

The question:
How many videos streams (assume HD resolution1080p) can each Panorama support for running AI inference, say, human detection?

Thanks in Advance,
Ari Telias, CTO
AiDANT.ai

Edited by: AriTelias on Feb 21, 2021 11:14 AM

demandé il y a 3 ans297 vues
2 réponses
1

Hi,

This is quite a challenging question: the Panorama appliance runs a Jetson AGX Xavier architecture, which is significantly different than RTX. For instance, it sports specific deep learning accelerators (DLA) that offload the GPU for deep learning inference workload, making it hard to compare the number of TensorCores or Cuda Cores. It also depends on the frame rate you are looking at. Also keep in mind that NVidia Jetson platform does use a lot less energy than standard grade RTX video cards.

At this stage, you would have to test your workload with a development kit (do reach out to AWS representative if you feel you need support to perform these tests). AWS is collecting more datapoints from customers using the appliance and from internal usage for benchmark purpose. The Panorama service team may have more insights about the framerate you could potentially achieve given your video frame rates, resolutions, model size and architecture.

--
Michaël Hoarau
AWS AI/ML Specialist Solutions Architect

répondu il y a 3 ans
0

Thank you!

répondu il y a 3 ans

Vous n'êtes pas connecté. Se connecter pour publier une réponse.

Une bonne réponse répond clairement à la question, contient des commentaires constructifs et encourage le développement professionnel de la personne qui pose la question.

Instructions pour répondre aux questions