DeepStream SDK on NVIDIA Jetson

What is DeepStream SDK?

Build and deploy AI-powered Intelligent Video Analytics apps and services. DeepStream offers a multi-platform scalable framework with TLS security to deploy on the edge and connect to any cloud.

• Develop in C/C++ or Python
• Built for scalability - one application for NVIDIA T4 and Jetson platforms
• Highest throughput for object detection, image classification and semantic segmentation models
• Deploy models in native DL frameworks such as Pytorch and TensorFlow for inference
• IoT integration interface with Kafka, MQTT and AMQP
• Turnkey integration with AWS IoT and Microsoft Azure IoT
• Multi-GPU, multi-stream and batching support for high throughput inference

Why do we use DeepStream SDK?

DeepStream SDK design for NVIDIA Jetson Family. So, For the best performance.

Install DeepStream SDK

https://developer.nvidia.com/deepstream-getting-started

For NVIDIA Jetson

https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_Quickstart.html#jetson-setup

Sample Python Applications

We currently provide the following sample applications:

• deepstream-test1 -- 4-class object detection pipeline
• deepstream-test2 -- 4-class object detection, tracking and attribute classification pipeline
• deepstream-test3 -- multi-stream pipeline performing 4-class object detection
• deepstream-test4 -- msgbroker for sending analytics results to the cloud
• deepstream-imagedata-multistream -- multi-stream pipeline with access to image buffers
• deepstream-ssd-parser -- SSD model inference via Triton server with output parsing in Python
• deepstream-test1-usbcam -- deepstream-test1 pipeline with USB camera input
• deepstream-test1-rtsp-out -- deepstream-test1 pipeline with RTSP output

Detailed application information is provided in each application's subdirectory under apps.

deepstream-test1 python 

In this sample, we first create one instance of "nvinfer", referred as the pgie.
This is our 4 class detector and it detects for "Vehicle , RoadSign, TwoWheeler,
Person".

Run Linux Command

jetson@nano:~/deepstream_python_apps/apps/deepstream-test1$ python deepstream_test_1.py /home/jetson/Videos/streams/sample_720p.h264 

TEST on NVIDIA Jetson Nano

• Jetpack SDK 4.5.1 • DeepStream SDK 5.1 • Python 3.6.9 • CUDA 10.2.89 • TensorRT 7.1.3 • Machine Learning Model : ResNet10 backbone ( FP16 ) • Video 1200x720 H.264

Python Source code

https://github.com/NVIDIA-AI-IOT/deepstream_python_apps

Performance Compare

Adun Nantakaew อดุลย์ นันทะแก้ว 081-6452400

LINE : adunnan

FaceBook : https://www.facebook.com/softpowergroup

Youtube : https://www.youtube.com/channel/UCw3VVy4wOsb8a0a1YYqJAPg

Web Blog : http://raspberrypi4u.blogspot.com/

Website : https://softpower.tech