How-To: Set up NVIDIA Jetson Nano for Object Recognition

by Miguel Alatorre, ameriDroid Technician:

In the eyes of many, NVIDIA is focused solely on graphics cards; this, however, is not entirely true. NVIDIA also has a hand in the artificial intelligence (AI) sector. They make sure hardware has the power to drive AI in data centers all the way to self driving cars; they go as far as to say AI “drives us today.” Keeping up with that message, NVIDIA makes it a goal to get AI in the hands of as many as possible with various boards, such as the Jetson Nano. Because of the Jetson Nano’s reach, it is inevitable that it will be a first for many; as such, NVIDIA has created a page on GitHub to walk users through the process of starting an AI project.

To start off, the user needs to prepare a microSD card with an OS flashed on it; we advise using a card larger than 16GB as the files for operation are quite large. The OS file can be found at https://developer.nvidia.com/jetson-nano-sd-card-image-r322, and balenaEtcher, our preferred program, can be found at https://www.balena.io/etcher/. To prepare the SD card, all that needs to be done is to plug it into your computer -- if needed you can find our SD adapter here. From here, open balenaEtcher and choose the .img file that came from the downloaded .zip file. Next, if not done automatically, select your SD card in the “target” field. From here, hit flash and let it run. Finally, install the card in the section underneath the main module.

Upon first boot, continue through the OS installer like normal. Once on the desktop, open the terminal with Ctrl + Alt + T and run the following commands:

sudo apt update && sudo apt upgrade

From here, follow NVIDIA’s Hello AI World tutorial starting on “Building the Project from Source.” To quickly get a live camera demo, the “Coding Your Own...” sections can be skipped, as they are for further development of the base demo. From here, you can continue to develop the scripts or train the AI. You’ve now finished preparing the Hello AI World tutorial!

NOTE: For live camera demonstrations, cameras like the Raspberry Pi Camera module are required.

Commands to remember:

• ImageNet
• Console
• //    “network=” is optional
• C++: ./imagenet-console --network=NETWORK INPUT.jpg OUTPUT.jpg
• Python: ./imagenet-console.py --network=NETWORK INPUT.jpg OUTPUT.jpg
• Camera
• //      “network=” through “height=” are all optional
• C++: ./imagenet-camera –network=NETWORK –camera=/CAMERA/PATH –width=WIDTH --height=HEIGHT
• Python: ./imagenet-camera.py –network=NETWORK –camera=/CAMERA/PATH –width=WIDTH --height=HEIGHT
• DetectNet
• Console
• C++: ./detectnet-console --network=NETWORK INPUT.jpg OUTPUT.jpg
• Python: ./detectnet-console.py --network=NETWORK INPUT.jpg OUTPUT.jpg
• Camera
• //      “network=” through “height=” are all optional
• C++: ./detectnet-camera –network=NETWORK –camera=/CAMERA/PATH –width=WIDTH --height=HEIGHT
• Python: ./detectnet-camera.py –network=NETWORK –camera=/CAMERA/PATH –width=WIDTH –height=HEIGHT

Our sample photos:

For more product updates and How-Tos, follow us on YouTube, our ameriBlog, Facebook, Twitter and Instagram!