Do it yourself ai:google aiy Developer Kit parsing

At the 2018 International Consumer Electronics Show (CES), one of the most obvious trends is the presence of Amazon and Google's voice technology, such as Amazonalexa stationed in the Acer pen, Google assist into the Kia car, other such as smart TVs, smart speakers, Even the traditional digital recorder TiVo has become a preemptive target.

google Assistant

speech recognition in battle

This is a new wave of battle, in the past, Google has the advantage of the search portal for PC Internet, so that Google earns a huge advertising agency revenue. But the form of the Internet in the change, including mobile phone "say", living room lying when the wisdom horn "say", or drive to the computer "said", all to get a search response, and voice response, which became the beginning of the speech recognition in the war, and more intense than the past search engine. Search engines can also appear 三、五个 results options, but the voice response can only be read one by one, listening to the first three probably no patience, so the answer to the first response will be more important.

in order to make our own voice recognition, cognitive computing, artificial intelligence and other technologies more popular, Amazon and Google also launched a related suite, and as far as possible to provide the most affordable to interested developers, For example, June 2016 Amazon is based on the Raspberry Pi computer (RPi 2 Model B) with its Alexa software, you can module out the effect of Amazon Echo, and related development (more information can refer to here).

https://aws.amazon.com/cn/blogs/china/raspberry-alexa/

google intelligent Voice, strong attack

voice Kit

Last year, Google also launched a similar kit, called Aiy, taken from AI (Artificial Intelligence), with its own DIY (do Ityourself) is a combination of two words, meaning a DIY ai kit. Google pioneered the voice version of the suite, the Google AIY Voice kit, and then launched the visual version of the suite Google AIY Vision Kit.

google Aiy may be inspired by Amazon, so also based on the Raspberry Pi computer, and then catch the necessary hardware peripherals and parts, the most basic needs of the voice, visual identification development device, It may also be affected by the IBM-initiated Tjbot project (also a raspberry-based speech recognition, AI appliance), with the addition of a cardboard shell and a simple interactive interface, such as a three-color LED light and a button, but no Tjbot swing arm (driven by a servo motor).

google launched the first voice version (known as V1), the use of the Raspberry Pi computer is still generally the most popular model B (Raspberry Pi 3), the price is higher, about 35 U.S. dollars. However, the later version of the Voice and visual version of the more inexpensive Pi Zero W (wireless version of Pi Zero), the price of about 10 U.S. dollars.

Regardless of whether the visual or voice suite contains Raspberry Pi, it also contains the microSD memory card required by the Raspberry Pi, and if the visual version also contains the Raspberry Pi Camera module (V2 version), The price for this voice edition kit is about $49.99 and the visual version is $89.99.

Visual  VS  Voice Edition

Since there is a difference between the visual version and the voice version, where is the difference? In fact, the visual version and the voice version have some of the same accessories, such as three-color LED lights, buttons, Gpio pin line, etc., two version of the main difference is: The visual version of the interface Board card (known as Visionbonnet Board, voice version is called Voicebonnet Board) more than one image processing AI chip, also known as VPU (Vision Processing Unit visual processing Unit), the Intel Movidius Ma2450,movidius is intel in 2016 to buy and the industry.

visual version

ma2450 can enhance the visual operation, reduce the visual computing load of the Raspberry Pi computer, and also perform the operation of the neural network, which is the main reason that the visual version is worth more than $40. Of course! The visual version also provides a camera and camera cover mirror, but there is no voice version of the speakers, instead of only a simple buzzer.

voice version

Because the voice operation is significantly lighter than the visual, pure consumption of the Raspberry Pi computing power can be, do not need an additional accelerator chip, so costs and pricing can be lower than the visual version.

vision kit

It is worth noting that the visual version has three sets of TensorFlow-based neural network model software: The first set is the disk of the object identification, is the open source code of the Neural network model project Mobilenets built, Can identify thousands of common daily necessities; the second set of recognizable images of the face, and through facial expressions to analyze the current mood of pleasure, to play a happy score; the third set can identify which of the images in the human, what is the cat, what is the dog.

Moreover, there are other identification models listed on the website, such as: What do you recognize in the dish? Most models are based on mobilenets, but there are also models based on Squeezenet to identify the natural landscapes.

Visual version is also included in the compilation tool software, can be on the workstation or the cloud, the model for training, retraining and other compilations. As for the voice version, the accompanying software is mainly Google Assist's software Development Kit (SDK) and demo application (demo application).

Other packages also include the Google Play App Store on the rack to find Google AIY projects's mobile app (APP), available for free download and installation, and developed with Aiy hardware suite.

concluding remarks

google at the end of a number of issues, so as to stimulate the development of artificial intelligence to write power, for example, with the visual version of the development of a variety of animal and plant AI software to identify whether the car is running off the lane? Identifying your visitors ' satisfaction with your family's festive decorations is expected to be completed and implemented faster, based on the software and hardware of the visual suite.

original reference

https:// mp.weixin.qq.com/s?__biz=mzu0mtg0odizna==&mid=2247484725&idx=1&sn=6d53e71d6a7f0b4b8aca0d2ab253cf6a &chksm=fb22e2f0cc556be61a07a93a7ce2221d4a866eaeb6ad124c06f7f5dd6051a138c77ab692d524#rd

Do it yourself ai:google aiy Developer Kit parsing