NEXT-SYSTEM Implements Their Pose Estimation AI Engine “VisionPose” Into Renesas’ Image Recognition AI Microprocessor
Advanced Pose Estimation Through Edge-Side IoT Devices – Expanding Supported Platforms for Development in the Edge AI Field.
FUKUOKA CITY, FUKUOKA PREFECTURE, JAPAN, November 30, 2021 — NEXT-SYSTEM implemented their independently developed pose estimation AI engine “VisionPose” into Renesas Electronics Corporation’s RZ/V2M microprocessor for AI image recognition and develops an area intrusion detection application that sends out alert notifications when a specific body part of a person enters a designated area.
The Vision AI MPU is a semiconductor that has functions to speed up arithmetic processing with mechanisms to perform machine learning and deep learning. By using Renesas’ Vision AI MPU, which combines high-speed AI inference with low power consumption, it is possible to run the pose estimation AI engine, which requires a huge amount of computational processing on the edge side. Now that VisionPose can run on this MPU, various IoT devices equipped with it will be able to perform advanced AI processing such as pose estimation and behavior analysis.
Simplifying The Use of Vision AI MPUs in IoT Devices by Implementing Pose Estimation Technology
In general, AI processing requires a huge amount of computational processing compared to conventional software. When using AI for embedded applications, conventional solutions that use CPUs and GPUs are difficult to use due to their high-power consumption, heat generation, and limited installation space. VisionPose is no exception, and in order to perform pose estimation on IoT devices and edge devices, the corresponding hardware is essential. Amidst these challenges, the Renesas Vision AI MPU “RZ/V2M” has emerged as an MPU with a built-in AI accelerator that offers both high performance and low power consumption as well as the flexibility to respond to advancement.
The AI accelerator ‘DRP-AI’ equipped with this MPU is a dynamic reconfiguration processor originally developed by Renesas to provide flexibility, allowing DRP-AI alone to perform not only AI inference processing but also pre- and post-processing, which is essential for image AI, while achieving high power efficiency. NEXT-SYSTEM implemented the pose estimation AI “VisionPose” into Renesas’ Vision AI MPU, and as a first project developed an area intrusion detection application. The expansion of platforms with Renesas’ Vision AI MPUs has made it easier to support VisionPose in embedded products in the IoT and edge AI fields. This makes it easier for more customers to use this pose estimation technology.
What is VisionPose?
VisionPose is an AI engine that detects human skeleton information on camera images, still images and videos, without the need for markers or special equipment. As an all-around SDK package, it can be applied to various fields in development and research. With the “Standard” edition of VisionPose and the use of 2 cameras, high-precision skeleton detection in 3D coordinates can be performed, and by using the VisionPose “Single 3D” edition, skeleton detection with only one camera is possible. With this, NEXT-SYSTEM provides two SDK packages for all-around purposes. Sales outside of Japan have started with two products, VisionPose Standard Windows C# and C++, but the number of supported platforms will be expanded sequentially.
What is the Demo App “Area Intrusion Detection”
Through implementing VisionPose, this application detects and sends out a notification when a specific body part (such as hand or leg) of a person enters a specified area. Different from object detection technology, which only roughly captures objects, pose estimation technology makes judgments based on the skeletal information of a person, thus detailed settings can be configured for each part of the human body.
Comment from Renesas Electronics Corporation Senior Executive Manager, Mr. Shigeki Kato (Enterprise Infra Solution Division)
“We are pleased to announce that our Vision AI MPU “RZ/V2M” has been equipped with “VisionPose”, a pose estimation AI engine originally developed by NEXT-SYSTEM. By combining “RZ/V2M”, which has high power efficiency when running AI, and “VisionPose”, our Vision AI MPU can be applied to any IoT devices regardless of their installation location. We expect that this will enable our customers to accelerate the development of products, such as anomaly detection systems, and contribute to the future spread of edge AI.”
About NEXT-SYSTEM Co., Ltd.
NEXT-SYSTEM is a Japanese IT company founded in Fukuoka City in 2002, and since then has been focused on the research of behavior analysis through AI technology, ergonomic system development and research, development of cutting-edge systems, such as xR (AR/VR/MR), and the development and sales of their Pose Estimation AI Engine “VisionPose” and AR Signage System “Kinesys”.