Jetson Nano is popular for its AI capabilities in a small form factor and enabling numerous AI applications. But it also packs powerful HW Codecs for H.264/H.265 format and supports encoding and decoding up to Ultra HD resolutions. Unlike other Jetson devices, Nano has a port of ffmpeg that uses HW codecs. Availability of ffmpeg on Nano makes it readily integrate with VideoCoin Network, and provides a very power efficient VideoCoin Worker.

Setup

Follow the usual setup of downloading Ubuntu 18.04 64 bit OS and flash the microSD card as per the instruction specified in the following NVidia's developer site :

https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit#write

Download and setup Ffmpeg for Nano

The following git repo contains a patch for enabling FFMPEG to use on Jetson Nano.

https://github.com/jocover/jetson-ffmpeg.git

Follow the instructions in the site and prepare ffmpeg to use Nano HW Codecs,

Download the VideoCoin Worker Docker Image or Binaries for Nano

You can learn more about setting up a worker here​

Setup VideoCoin Worker in VideoCoin Network Console

Use the "Worker" tab in VideoCoin Network Console and create the worker. The "Client Id" assigned to the worker needs to be used while running the worker as described in the following section.

Start the worker

Set the environment variables and start the transcode worker as shown below.

Environment variables and command-line options

Performance:

Nvidia jetson-nano supports encoding of four streams at 1080p@30fps H264 or H265.

We tested the worker for FullHD resolution at 30fps and following graph shows CPU usage when jetson-nano

The above graph can be interpreted as the VideCoin worker is handling FullHD transcoding with live streaming without any hiccups.

Jetson nano can be used to transcode up to 4k @30fps resolutions.