Waveshare Jetson Nano Developer Kit Small Computer for AI Development Run Multiple Neural Networks in Parallel for Image Classification Object Detection Segmentation Speech Processing

The Jetson Nano as you see on Nvidia's website. Waveshare's customer service has been the best I have received from any company. They have reached out to me on every order to change over to the UK warehouse for Prime - this is my mistake. Because of this, I went direct to Waveshare's website and ordered some other parts we needed.

Connected a 5V 4amp power source from an older SBC. Connected a 4cm fan to the beast heatsink, as gets toasty even without utilising the CUDA cores. Connected a RaspPi camera to the CSI port. Tried a bit of facial recognition following the Jetson Hacks tutorial on YT. The GPIO don't use the same RaspPi standard power delivery is different, so won't be using them as extra components are required. Intending to try various AI guides. Overall fun to play with.

Couple of tips: Go for a larger Micro SD, e.g. 32GB or more. Definitely get the 5v 4000mA PSU and don't muck about with trying to power it over the USB. Also recommend the (fairly expensive) metal case to protect it. Overall a great product - as you would expect from NVidia.

The system is designed to support AI with Deep Learning on an Ubuntu Linux system. The hardware works straight out of the box. You can download an Ubuntu 18.04 image from the Nvidia site (see little leaflet that comes with the device). For that you need to provide a reasonably large (e.g. 64 GB) SD card which fits neatly into a SD card reader slot on the board. Do not try to pull the SD card out by force! Press slightly and it should pop out. Flashing the SD card is standard procedure. Follow the instructions on the little leaflet that comes with the box. Do not forget to purchase an appropriate WiFi adapter (maybe try it on a basic RPi first to make sure it actually works in your environment). I chose to use also a wireless keyboard and mouse. Note that the device is very light and the electronics are not protected through a case when they come out of the box. I recommend using an adequate power supply for the system and short USB cables. The Jetson Nano is very power hungry indeed and will stall or simply shut down if the power supply is not sufficient. The Linux image works out of the box like a charm. Connect to the Internet and you are ready to roll, i.e. apt-get update and upgrade as usual. Please note that the LINUX system is really quite basic and you need apt install and pip install a lot of software before you can run your first Keras, Theano or Tensorflow GPU supported Deep Learning example. Depending on the Deep Learning application, the accelerations through the GPU can be significant though. I am comparing with my trusty Odroid N2 CPU based system which is for a SoC quite nimble already; however, the Jetson Nano using the GPU achieves between 10 to 100 times faster results on my NMIST tests where GPU support is critical. The Jetson is certainly useful for developing and practical tests. The portability of the small Jetson Nano makes it useful for Deep Learning image processing tasks (e.g. on autonomous vehicles) where you need quick GPU based model evaluation. However, remember that the Jetson requires sufficient power to run.