Simulation / Modeling / Design

NVIDIA Releases Latest JetPack 3.1 SDK with TensorRT 2.1 for AI at the Edge

Jul 24, 2017
By Brad Nemire
Discuss (0)

AI-Generated Summary
  • TensorRT 2.1 doubles deep learning inference performance for low latency batch size 1 applications on Jetson TX1 and TX2 platforms.
  • JetPack 3.1 provides a unified code base with shared Linux kernel 4.4, libraries, and APIs for both TX1 and TX2 devices.
  • The release includes CUDA Toolkit 8.0.82, cuDNN 6.0, VisionWorks 1.6, Tegra System Profiler 3.8, and Tegra Graphics Debugger 2.4 for robotics and intelligent machine development.

Next Steps

  • Download JetPack 3.1 to begin developing for Jetson platforms.
  • Build a demo in two days using the provided quick start guides and pre-trained network models.
  • Watch the on-demand webinar covering tools and techniques for autonomous robot development.
Powered by NVIDIA Nemotron. AI-generated content may summarize information incompletely. Verify important information. Learn more

NVIDIA released JetPack 3.1, the production software release for the Jetson TX1/TX2 platforms for AI at the edge. JetPack 3.1 doubles the deep learning inference performance for low latency applications for batch size 1. For the first time, developers will have access to the same unified code base for both TX1 and TX2, including a shared Linux kernel 4.4 codeline, libraries and APIs. This comprehensive package gives developers advanced tools to create tomorrow’s robots and intelligent machines.

Figure 1: Inference throughput of GoogLeNet and ResNet-50 using Jetson TX2 Max-Q and Max-P power profiles. TensorRT 2.1 provides twice the inference throughput on GoogLeNet and ResNet.

The Jetson platform has already been deployed across a variety of applications including drones, industrial and delivery robots, human support robots (HSR), high school robotics programs, telepresence, video analytics, and more. Now with the release of JetPack 3.1, large deep learning models can be run with improved performance using the latest tools and libraries. They include:
JetPack 3.1 is available now and free to download. Developers can start creating today with a variety of design resources available. They include:

NVIDIA JetPack 3.1 — Software Components
Linux For Tegra R28.1 Ubuntu 16.04 LTS aarch64
CUDA Toolkit 8.0.82 cuDNN 6.0
TensorRT 2.1 GA GStreamer 1.8.2
VisionWorks 1.6 OpenCV4Tegra 2.4.13-17
Tegra System Profiler 3.8 Tegra Graphics Debugger 2.4
Tegra Multimedia API V4L2 Camera/Codec API

Two Days to a Demo
Dive into deep learning with NVIDIA’s guide to building a demo in two days. All the necessary tools are provided, including quick start guides, software samples and pre-trained network models.
On-Demand Jetson Webinar
Listen to NVIDIA’s on-demand Webinar on “Breaking New Frontiers in Robotics and Edge Computing with AI.” It covers the latest tools and techniques to develop autonomous robots and other intelligent machines with NVIDIA Jetson TX2.
Learn more about what NVIDIA is doing for the future of robotics with the Isaac Initiative, including the NVIDIA Jetson Reference Platforms and Isaac Lab.
Read our Developer blog for a deeper dive into our latest JetPack 3.1 release >

Discuss (0)

Tags

Computer Vision / Video Analytics | Robotics | Simulation / Modeling / Design | CUDA | cuDNN | Jetson | News | Machine Learning & Artificial Intelligence

About the Authors

Avatar photo
About Brad Nemire
Brad Nemire leads the Developer Communications team at NVIDIA. Prior to NVIDIA, he worked at Arm on the Developer Relations team. Brad graduated from San Diego State University and currently resides in Silicon Valley.

Comments