If artificial intelligence is to be popularized, this step must be taken first!

Thanks to the rapid development of software and hardware, artificial intelligence, which has been around for decades, has finally experienced a big explosion in recent years. Especially on the terminal side, the growth rate of artificial intelligence driven by various applications is astonishing.

According to HIS data, the market demand for edge AI began to explode in 2018, and the market size increased from US$400 million in 2017 to US$1.9 billion in 2018. The growth rate is expected to exceed 400% in 2019, with more AI applications and products being deployed on the "edge" of the network to achieve an AI technology application path with lower latency, lower energy consumption, smaller size and lower cost.

But we should see that behind these manufacturers rushing into this market, there are still many problems waiting for developers to solve.

Taking the development of an intelligent surveillance camera for automatic identification of vehicle license plates in and out of parking lots as an example, developers must not only be concerned about system development costs, but also software and hardware costs, as well as the costs of subsequent maintenance and data security issues. In addition, development issues including chips, memory, networks, bandwidth, and various device specifications need to be considered. Other issues such as data preprocessing, model training, effective management and maintenance of massive devices, and effective handling of personal privacy and security issues will become "stumbling blocks" for developers to implement end-side AI products. That means a one-stop solution to all problems has become an urgent need for developers.

Yang Yuxin, vice president of Thundersoft, also pointed out in an interview with Semiconductor Industry Observer that current AI developers face two problems: first, how to find scenarios; second, artificial intelligence will be a cloud- and end-oriented solution in the future, especially end-side artificial intelligence, which will become mainstream in the future, but it is difficult for current developers to find an embedded or terminal-side artificial intelligence development platform with good performance and mature platform.

To this end, ThunderSoft has developed the world's first AI development kit, Thundercomm TurboX AI Kit, based on Qualcomm Snapdragon 845 mobile platform to help developers solve the problem of edge AI development. "Qualcomm products are very mature, with excellent system stability, functional integrity and AI capabilities. This combined with our software and system strengths can bring developers a better experience," added Yang Yuxin.

According to the data, Qualcomm Snapdragon 845 platform is its third-generation artificial intelligence (AI) platform. It relies on its Hexagon 685 DSP, Adreno 630 GPU, Kryo 385 CPU, Snapdragon Neural Processing Engine (NPE) SDK, Android Neural Network (NN) and Hexagon NN and other hardware and software. Snapdragon 845 enables developers to achieve better overall AI performance on the terminal, regardless of network connection.

Since its establishment in 2008, ThunderSoft has been engaged in the research and development and innovation of operating system technologies such as Android, Linux, Windows and RTOS. Years of development have enabled the company to form a comprehensive technical system from hardware drivers, operating system kernels, middleware to upper-level applications. It has accumulated rich R&D experience and many proprietary intellectual property rights, and has strong technical development and service capabilities. The core technologies cover communication protocol stacks, deep learning, graphics and image algorithms, operating system optimization, and security technologies. On this basis, the company has turned its attention to artificial intelligence technology to help and accelerate productization and technological innovation in the fields of smart phones, smart Internet of Things, and smart cars.

From Yang Yuxin's introduction, we learned that in terms of hardware, the AI ​​Kit launched by Thundersoft has rich interfaces including microSD, micro HDMI, Ethernet interface, 1 USB Type C and three USB ports. It also has a built-in Ultra HD camera, which can be expanded to USB camera, IP camera, microphone and speaker, etc., allowing developers and manufacturers to flexibly choose the visual solutions they need.

In terms of software, the TurboX AI Kit running on Android O integrates the Qualcomm Snapdragon Neural Processing Engine (SNPE) designed for running neural networks on the terminal; it is pre-installed with an AI algorithm software development kit (SDK) optimized based on the Qualcomm artificial intelligence engine, including face recognition and object recognition SDKs; at the same time, it can use open source artificial intelligence services such as Google TensorFlow and Facebook Inc.'s Caffe/Caffe2 and Android NN API; in addition, it also supports a variety of cloud computing service platforms including AWS video kinesis streaming and Microsoft Azure, as well as AI Studio and API Cloud development.

Such a powerful and ecologically rich AI development platform can provide sufficient support for developers to evaluate, develop, optimize and verify end-side artificial intelligence. For developers who have in-depth knowledge in the field of algorithms, AI Kit can quickly implement their algorithms, easily connect to the Android ecosystem, and expand their applications; for module manufacturers, AI Kit can also provide them with the opportunity to quickly build product prototypes.

Yang Yuxin also emphasized that for ThunderSoft, the new platform can help them accumulate more applications and algorithms, demonstrate ThunderSoft's strength in AI software and hardware, and provide extensive support for developers. This will also provide sufficient support for ThunderSoft's three main businesses of "intelligent software systems, smart cars and smart Internet of Things".

He further pointed out that ThunderSoft will continue to upgrade AI Kit from both software and hardware aspects in the future, bringing more functions to it, and continue to bring better experience to developers, providing them with sufficient motivation and support to develop new product systems and verify new prototypes. This is what ThunderSoft has always pursued.

In my opinion, lowering the development threshold to the lowest level is also a prerequisite for the explosion of applications. Let us wait together for the arrival of the golden age of artificial intelligence applications.