Q&A for the GTC webinar “A21337 Implementing Real-time Vision AI Apps Using NVIDIA DeepStream SDK”

Thank you for attending the GTC webinar “A21337 Implementing Real-time Vision AI Apps Using NVIDIA DeepStream SDK ”, presented by Amulya Vishwanath. We hope you found it informative. We received a lot of great questions at the end and weren’t able to respond to all of them. We are consolidating all follow-up questions in the following post.

  1. Can you provide a detailed document or any end to end demo for DeepStream with Azure IOT for multiple streams at a time?

Edge-to-cloud messaging is a built-in deepstream-test5 app. Here’s test5 documentation - https://docs.nvidia.com/metropolis/deepstream/dev-guide/DeepStream_Development_Guide/deepstream_reference_app_test5.html#. To learn more about Azure adaptor library please see the plugins manual -https://docs.nvidia.com/metropolis/deepstream/dev-guide/index.html#page/DeepStream%20Plugins%20Development%20Guide/deepstream_plugin_details.html#wwpID0E0IJ0HA

  1. If we have multiple devices, do we need to deploy multiple dockers at Azure?

Yes, each edge device will get it’s own DeepStream container

  1. Please show me h 264 r t s p camera reading and decoding pipeline into app sink R G B output

Here are few face detection models that will work with DeepStream: PeopleNet - https://ngc.nvidia.com/catalog/models/nvidia:tlt_peoplenet and FaceDetectIR - https://ngc.nvidia.com/catalog/models/nvidia:tlt_facedetectir

  1. What is DLA vs GPU in Jetson specs? Why is DLA so much slower?

DLA is a separate accelerator for AI called Deep Learning accelerator. DLA is only available on Jetson AGX Xavier and Xavier NX. GPU is more general purpose and is available on on NVIDIA Jetsons

  1. Can I get a pointer for K8 deployment for Jetson

Deploying AI apps with K8s and EGX on Jetson Xavier NX - https://developer.nvidia.com/blog/deploying-ai-apps-with-egx-on-jetson-xavier-nx-microservers/

  1. Can we combine OpenCV with Deepstream? for preprocessing purpose for example

Yes, check out documentation on implementing custom GST plugin for OpenCV with DeepStream - https://docs.nvidia.com/metropolis/deepstream/dev-guide/DeepStream_Development_Guide/deepstream_custom_plugin.html#

  1. Are there security features for the transfer of the models

Deepstream supports models encrypted using TLT, which would allow transfer of models over insecure channels without getting compromised. Preferably, models should be transferred using encryption (eg: using HTTPS); as is the case when downloading models from NGC.

You can watch the recording and download the presentation slides from the following link

For the response to the follow-up questions for another GTC talk “A21333-Accelerating Vision AI Applications Using NVIDIA Transfer Learning Toolkit and Pre-Trained Models”, please visit TLT forum.

If you have more questions, please feel free to post your questions in the forum and we will further assist you