Dev Kit Weekly: SmartCow Apollo Audio/Visual AI Engineering Kit

August 19, 2022

Video

The Apollo Audio/Visual AI Engineering Kit includes all the hardware and software you’ll need to develop* and deploy your own vision, audio, or conversational AI application. At just 81 x 69 x 125 mm, it contains four microphones, two speaker terminals, an 8MP camera module, 128GB NVMe SSD, and an NVIDIA Jetson Xavier NX compute module that delivers as much as 21 TOPS of performance.

NVIDIA’s Jetson Xavier NX compute module is an energy-efficient edge processing platform capable of running multiple neural networks simultaneously while consuming just 10, 15, or 20 Watts of power. That PPW is possible thanks to a 6-core 64-bit ARM®v8.2 CPU with 6MB of L2 and 4MB L3 cache. Of course, that runs alongside an NVIDIA Volta™ GPU with 384 NVIDIA CUDA® cores and 48 Tensor cores, optimized for workloads like conversational AI applications.

Voice data captured by four onboard MEMS microphones or over the Apollo’s 3.5 mm line-in or I2S interface ends up on the Jetson Xavier NX module where signal processing, filtering, and algorithm execution is performed. It’s also where video and images streaming from the included 8MP Sony IMX179 CMOS image sensor camera module or in from the MIPI CSI-II lanes, accessible via the kit’s 15-pin front-panel connector, are processed as well.

As it moves from peripherals to the CPU and on to Tensor and CUDA cores, then eventually to the cloud or one of the kit’s two speaker terminals, 3.5 mm line-out jack, or other output, audio and multimedia data is transferred to and from memory multiple times. To ensure there’s plenty of space to read and write all that signal data, the NVIDIA Jetson Xavier NX module includes 8GB of 128-bit LPDDR4 memory that runs at 1600MHz and is capable of data transfer speeds as fast as 51.2 GB/s; 16 GB of additional LPDDR4 is also present, only this runs at 59.7 GB/s. There’s also 16GB of eMMC 5.1 storage and a 128 GB M.2 NVMe SSD for data logging, etc.

To access that data and program space, developers can make use of two different USB ports. A variety of other connections that support multimedia application development can also be found on the Apollo, such as a mini DisplayPort interface and GbE via an RJ45 connector. There’s even this very noticeable 2-inch OLED status display.

Many of the hardware features we just mentioned pair with several supported SDKs that specifically target conversational AI development by offering capabilities like natural language processing (NLP), automatic speech recognition (ASR), and more.

For example, the Natural Language Toolkit, or NLTK, is a free and open-source Python SDK for computational linguistics that gives Apollo users API access to more than 50 corpora and lexical resources such as WordNet. These well-documented APIs assist in the creation of programs that analyze and interpret human language for translation, sentiment identification, entity classification, and so on.

Apollo also comes pre-packaged with:

• The NVIDIA JetPack SDK that’s flush with deep learning libraries
• NVIDIA DeepStream that emphasizes computer vision
• And the Riva SDK that provides an end-to-end ASR pipeline complete with customizable language and acoustic models.

Riva also includes a feature extractor, decoder, spectrogram generator, and vocoder models in support of a text-to-speech feature that can be used to create custom voices for your very own version of Alexa or Siri.

For NLTK, Riva, and other audio applications, Apollo’s included software tutorials walk you through connecting and configuring your microphones and speaker, then utilizing the PyAudio library to control incoming and outgoing audio signals.

So, now you know what you need to BYO (build-your-own) Echo, and you have all those tools at your fingertips with SmartCow’s Apollo Audio/Visual AI Engineering Kit.

If you’d like to get your own little yellow submarine, you can contact SmartCow’s sales team via email at [email protected], or click the “Buy Apollo” link on the apollo.smartcow.ai microsite and fill out the form. Price: Unknown.

But, I can tell you how much this one in the video costs: $0. That’s if you enter the raffle below to win this week’s kit. If you’re that lucky someone who wins, we’ll ship it to you anywhere in the world for free, too.

As always, thanks for watching, Godspeed in the raffle, and we’ll see you next week on Dev Kit Weekly.