First steps, any help is greatly appreciated

[rferrandis]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU)
GPU first, then Jetson for other applications
• DeepStream Version
6.3
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
536.99
• Issue Type( questions, new requirements, bugs)
Question
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

Hi,

I am working on a real time object detection application Initially, I tried with just the Triton server with a multi GPU setup, but the throughput is insufficient. I’m using an SSD Mobilenet v2 640 model without batching (I don’t know how to make a model like that perform with batching enabled, but that’s a different matter.)

I have a few questions and I hope that this is the right place to ask:

• Is it possible to run the doker image on a windows 10 computer? I’ve managed this with Triton, but I’m not sure if it will work with the Deepstream image, it seems to be more complex than the Triton one.
• Do I need to maintain a separate Triton service? Or does the deepstream docker image come with all necessary components?
• Regarding the cameras, can I apply preprocessing custom filters to the captured images? Or is the pipeline fixed?
• And the last one, I’m sending 640 tiles with some overlap to the inference server, Can I achieve this with the deepstream sdk? I don’t wish to resize the images due to the potential loss of quality, especially since the objects I need to detect are quite small.

Thank you in advance!

[Fiona.Chen]

@rferrandis

• Is it possible to run the doker image on a windows 10 computer? I’ve managed this with Triton, but I’m not sure if it will work with the Deepstream image, it seems to be more complex than the Triton one.

No. There is no docker for Windows.

@rferrandis

• Do I need to maintain a separate Triton service? Or does the deepstream docker image come with all necessary components?

It depends on your environment and requirement. The deepstream:6.3-triton-multiarch docker has included all Triton components needed by DeepStream. Some tools or some backends which are not used by DeepStream are not in the docker. DeepStream | NVIDIA NGC

@rferrandis

• Regarding the cameras, can I apply preprocessing custom filters to the captured images? Or is the pipeline fixed?

DeepStream is a SDK. It provides APIs and some sample applications to tell you how to use the APIs. There is no limitation for you to add your own customizations.

@rferrandis

• And the last one, I’m sending 640 tiles with some overlap to the inference server, Can I achieve this with the deepstream sdk? I don’t wish to resize the images due to the potential loss of quality, especially since the objects I need to detect are quite small.

The gst-nvinferserver is the DeepStream Triton Interfaces and it is open source. If you want any customization, please investigate the doc and code Gst-nvinferserver — DeepStream 6.3 Release documentation.