Edge AI Vision KitsOur Product Insights

How to run DeepStream applications out-of-the-box on SmarteCAM

NVIDIA® DeepStream Software Development Kit (SDK) is an accelerated AI framework for building Intelligent Video Analytics (IVA) pipeline. See what happens when we run an NVIDIA DeepStream pipeline on SmarteCAM and understand its use cases and performance.


e-con Systems’ SmarteCAM is a ready-to-deploy Artificial Intelligence-powered camera with powerful processing capabilities. It comes with an on-board Nvidia® Jetson™ TX2 CPU and a 256 core GPU to perform all the image processing and analytics indigenously. In this post, we will run an Nvidia DeepStream pipeline on SmarteCAM to show you the applications where this device can be deployed and how it performs.

What is DeepStream?

Nvidia® DeepStream Software Development Kit (SDK) is an accelerated AI framework to build Intelligent Video Analytics (IVA) pipelines. It runs on Nvidia® T4, Nvidia® Ampere and platforms, such as NVIDIA® Jetson™ Nano, Nvidia® Jetson AGX Xavier™, Nvidia® Jetson Xavier NX™, and Nvidia® Jetson™ TX1/TX2.

Basically, DeepStream is a bundle of plugins for the popular GStreamer framework. These plugins perform tasks typically needed for a deep learning video analysis pipeline. They are also highly optimized to run on a GPU. It results in a very efficient and quick video analysis pipeline since the images seldom leave the GPU.

One of the most interesting aspects of these plugins is the Gst-nvinfer plugin, which allows for running a TensorRT engine. In addition, there is also the Gst-nvtracker plugin for multi-object tracking and the Gst-nvstreammux plugin for effective multi-stream batching, among several others.


NVIDIA DeepStream pipeline is already installed in e-con Systems’ SmarteCAM. Once deployed, you are ready to go!

You can find the DeepStream folder here: /Opt/Nvidia/DeepStream/DeepStream-4.0

Let’s now run a sample object detection model.

Superior object detection with SmarteCAM

When we look at images or videos, we can easily locate and identify the objects of interest within seconds. Object detection is all about passing this intelligence to computers so that it can help locate and identify the object. It is leveraged across a wide variety of domains like video surveillance, image retrieval systems, autonomous driving vehicles, and others. While various algorithms can be used for object detection, we will be focusing on the YoloV3 algorithm.

Since SmarteCAM- IP66 rated camera with an HDR feature, it is perfect for outdoor applications – from detecting and identifying traffic violations to vehicle identification, license plate identification, crowd monitoring, etc.


Executing the sample application

To run the sample application, open the terminal and run the command

$ DeepStream-app -c /opt/Nvidia/DeepStream/DeepStream-4.0/samples/configs/DeepStream-app/source4_1080p_dec_infer-resnet_tracker_sgie_tiled_display_int8.txt


To run the sample application with our HDR camera, we need to do changes in the config file.

Move to the folder /opt/Nvidia/DeepStream/DeepStream-4.0/samples/configs/DeepStream-app/source4_1080p_dec_infer-resnet_tracker_sgie_tiled_display_int8.txt and do the changes in [source0] and [tracker] fields.

#Type – 1=CameraV4L2 2=URI 3=MultiURI 4=RTSP
#(0): memtype_device – Memory type Device
#(1): memtype_pinned – Memory type Host Pinned
#(2): memtype_unified – Memory type Unified

  • Open a terminal
  • Run the config file using DeepStream-app

$ DeepStream-app -c /opt/Nvidia/DeepStream/DeepStream-4.0/samples/configs/DeepStream-app/source4_1080p_dec_infer-resnet_tracker_sgie_tiled_display_int8.txt


To know more about the SmarteCAM, visit -> SmarteCAM or write to camerasolution@e-conststems.com

In our next blog, we shall dig deeper into SmarteCAM and better understand why it is the latest and fastest version of object detection. Stay tuned!

Related posts

Leave a Comment