“This is an introduction and actual implementation example of our low-end and high-end open edge AI platforms.” 

This is a seminar presentation titled “The Present and Future of On-Device AI.” 

On July 9, 2024, the 2nd Future Mobility Technology Seminar was held at the Suseong Hotel in Daegu, hosted by the Korean Society of Embedded Engineering. There were presentations from research institutes such as KETI, Infineon, GIST, and Yeungnam University under the title “Present and Future of On-Device AI,” and we were invited by the academic society to present on-device AI application cases in the CCTV field.

In this newsletter, we will share the materials presented at the seminar and provide a brief summary as follows.

Seyeon Tech has been continuously upgrading its hardware and operating system-independent software platform (middleware), and based on this, it supports everything from IP camera modules that operate on various SOCs and operating systems to Edge AI device management systems.

Currently, we are providing an open edge AI platform in three product lines, and the entry-level version has already been released and is being used by many customers. The high-end model is being developed for the end of August 2024, and the mid-range model is being developed for the end of December 2024. As shown in the table below, we are supporting customers to customize various products through NPUs of various speeds.

As shown in the picture below, customers only need to prepare a dataset and a learning server, and we support them in porting their learning model to our IP camera according to the guidelines we provide. Datasets or ONNX can be ported without external leakage, and the user can specify the learned model in the browser, providing an environment where AI operation and display are possible.

If necessary, the default libPostProces.so is provided so that users (customers) can directly implement the PostProcess step. We also support an interface that allows users to create and manage their own license keys, so you can rest assured about the security of the data sets and learning models you have developed.

We also provide an environment where customers can build or independently distribute OEM firmware including AI models developed by customers. Individuals or companies developing software are required to secure their own edge AI cameras so they can focus solely on software work, and various sensor devices linked to the edge cameras can be linked to enable various applications. By creating an ecosystem of an open edge AI platform, you can create a playground that allows flexible linkage and collaboration between rich AI models.

For further details, please download and check the presentation material below.

Presentation download link: https://flexwatch.notion.site/AI-8fa2edc3e0924a0e877bd2cef6d52c40?pvs=4

Seyeon Tech aims to provide an open AI camera and edge device development platform to development-based customers who want to utilize AI technology. By sharing our partners' core technologies, we aim to increase the utilization of cameras and edge devices.

If you need support to develop your own special AI Model, we can introduce you to companies specializing in AI Model.