This article is about using NVIDIA Jetson for computer vision applications. We will discuss what NVIDIA Jetson is, why it is popular for computer vision, and discuss the NVIDIA Edge AI devices suitable for deep learning tasks.
At viso.ai, we power the no-code computer vision platform Viso Suite. The Viso platform provides automated infrastructure and is used by organizations worldwide to develop, scale and operate computer vision applications.
Let’s jump right into the topic!
What is NVIDIA Jetson?
NVIDIA Jetson is a series of advanced embedding systems used by developers for creating innovative AI products across multiple industries. With Jetson being one of the world’s leading hardware platforms for AI on the edge, it is equally useful for students and tech enthusiasts to gain hands-on experience with AI through a range of creative projects.
The platform consists of small form factor, high-performance edge computers named modules. Additionally, it also includes JetPack SDK for software acceleration and a complete ecosystem to help speed up the development process of custom AI projects.
The NVIDIA Jetson system is high performance and power-efficient, making it one of the best and most popular platforms for building machines based on AI on the edge (Edge Machine Learning).
Why Is the Jetson Platform So Popular?
The NVIDIA Jetson systems come with numerous benefits that make it popular among professionals, software developers, and students alike. Some of its advantages include the following.
Modular Flexibility
Be it a small business or a large enterprise, the NVIDIA Jetson portfolio has modules perfect for each. You can choose from a range of modules perfect for developing everything between entry-level AI applications and advanced complex machines.
Unified Software
The NVIDIA Jetson is supported by unified software architectures, making the work of software developers much easier. The unified approach saves developers the hassle of repetitive coding when enhancing their creations on other Jetson modules.
The NVIDIA JetPack SDK comes with a Linux operating system (OS), CUDA-X accelerated libraries, and APIs for various fields of machine learning, including deep learning, computer vision, and more. It also supports machine learning frameworks like TensorFlow, Caffe, Keras, etc., as well as computer vision libraries like OpenCV.
Support For Cloud-Native Technologies
With support for cloud-native technologies and workflows (like orchestration and containerization), the NVIDIA Jetson platform offers developers the agility to rapidly develop or upgrade AI products.
The Latest NVIDIA Jetson Devices
Every NVIDIA Jetson module is a complete hardware package in itself. It comes with CPU, GPU, high-speed interfaces, memory, power management, and more. The Jetson devices are available in different power efficiency, performance, and form factor variations, making them suitable for use across all industries. The NVIDIA Jetson portfolio consists of the following edge devices or modules.
NVIDIA Jetson Nano
NVIDIA Jetson Nano is developed with a view of helping you add new capabilities to small AI systems. The module is perfect for students or developers just starting on their professional journeys as it is made for hands-on teaching and learning.
The module is small in size but high in performance and power efficiency – making it ideal for running modern AI workloads, multiple parallel neural networks, and processing data from high-resolution sensors all at once. As a result, the Jetson Nano is one of the most popular edge devices to create embedded AI solutions and products.
The NVIDIA Jetson Nano module is ideal for AI-based computer vision applications and to perform AI vision tasks like image classification, image segmentation, object detection, and more. It is compatible with open-source computer vision software and machine learning libraries like OpenCV.
The Jetson device is a developer kit that is accessible and comparatively easy to use. As a result, both beginners and experts use the jetson device with cameras for AI inference with deep learning. The kit is optimal for experimenting and creating a proof of concept (POC) of a next-gen AI solution. Comparing the Jetson Nano vs. Jetson TX2, the latter is significantly more powerful.
NVIDIA Jetson TX2 Series
The NVIDIA Jetson TX2 Series modules are built for exceptional power efficiency and speed despite being only the size of a credit card. Each of the modules features an NVIDIA Pascal GPU, 59.7 GB/s of memory bandwidth, up to 8 GB memory, along with a range of standard hardware interfaces.
The Jetson TX2 series provides up to 2.5 times the performance of the NVIDIA Jetson Nano module in as little as 7.5 watts. This allows you to efficiently run deep neural networks on edge devices achieving significantly higher accuracy. The modules available in the series include the Jetson TX2 NX, Jetson TX2, Jetson TX2 4GB, and the Jetson TX2i. Next, we will look at the differences between the TX2 series devices.
The NVIDIA Jetson TX2 NX shares pin and form-factor compatibility with the Jetson Nano and Jetson Xavier NX. The other three modules share the form-factor of the original Jetson TX2.
The usage of the TX2 series ranges in a variety of industries, including manufacturing, agriculture, retail, life sciences, etc. Of the four available modules, the Jetson TX2i is most suited for high-performance AI devices such as industrial robots, medical equipment, and machine vision cameras, due to its rugged design. For example, we have built an animal monitoring system with Computer Vision using the NVIDIA TX2, read our case study.
This makes those edge computing devices are suitable to power real-time processing where data is being generated (computing at the edge) due to the ability to operate with wide temperatures, wide voltage inputs, under shocks, and vibration.
Each module in the series comes with popular pre-trained AI models, the NVIDIA JetPack SDK, and the NVIDIA Transfer Learning Toolkit.
NVIDIA Jetson Xavier NX
At only 70 x 45 mm, the NVIDIA Jetson Xavier NX module is smaller than a credit card yet packs in high performance and power efficiency. This module was also said to be the smallest AI supercomputer in the world for embedded and edge systems.
This module is especially suited for running entire AI systems. With up to 21 TOPS of accelerated computing, it delivers the computing power needed to run parallel neural networks and process data from multiple high-resolution sensors simultaneously.
The Jetson Xavier NX also features support for cloud-native technologies, making it easier for developers to upgrade and manage the AI products. In addition, it comes with pre-trained AI models from NVIDIA NGC and the NVIDIA Transfer Learning Toolkit that allow for the faster creation of professional and optimized AI networks.
The NVIDIA Jetson Xavier NX is production-ready and comes with support for all popular AI models and frameworks. The module perfectly suits high-performance AI systems such as AI-powered high-resolution vision systems, smart cameras, drones, medical equipment, or commercial robots.
NVIDIA Jetson AGX Xavier Series
The NVIDIA Jetson AGX Xavier Series is explicitly developed for next-gen fully autonomous intelligent machines. Apart from the Jetson AGX Xavier Developer Kit, the series consists of the Jetson AGX Xavier and the Jetson AGX Xavier Industrial modules.
The Jetson AGX Xavier module delivers up to 32 TOPS of AI performance while using as little as 10 watts. It allows software developers to quickly deploy neural networks using a set of NVIDIA’s AI software tools.
The NVIDIA Jetson AGX Xavier Industrial is a rugged system on a module (SOM) that is optimized for high efficiency and performance. At around 20x of the performance of Jetson TX2i, it boasts of being the highest performing module for AI embedded systems.
It offers multiple specifications and functional safety capabilities, making it perfect for developers interested in creating industrial or safety-certified products. The high performance of the modules makes them suitable for several autonomous machines, including industrial and logistic robots, large drones, and factory systems.
What’s next?
The NVIDIA Jetson devices and modules are optimized edge devices to develop Edge AI systems for Computer Vision and ML at the edge.
To read more about related topics, check out the following articles:
