TL;DR:
- AI chips are set to generate $53.4 billion in revenue for the semiconductor industry in 2023.
- Himax Technologies leads with WiseEye2, an ultralow-power AI processor.
- WiseEye2 offers 32 times faster inference speed and enhanced precision.
- Embedded security and privacy features safeguard user data.
- CES 2024 will showcase Himax’s latest innovations, including automotive applications.
Main AI News:
The year 2023 witnessed an unprecedented surge in AI-powered applications across various industries, driving significant progress in data centers, edge infrastructure, and endpoint devices. This trend has not only revolutionized technology but also created a substantial revenue opportunity for the semiconductor industry. According to Gartner Inc., AI chips are projected to contribute a staggering $53.4 billion in revenue to the semiconductor sector in 2023, marking a remarkable 21% increase from the previous year. Furthermore, this growth trajectory is expected to continue, with forecasts reaching an impressive $67.1 billion in 2024 and soaring to an astounding $119.4 billion by 2027.
However, despite these promising forecasts, the AI semiconductor sector faces persistent challenges, particularly in terms of performance and power efficiency. To overcome these obstacles, concerted efforts are required to enhance and refine designs that facilitate the low-power execution of AI-based workloads.
Addressing the growing demand for AI applications in endpoint devices, Mark Chen, Vice President of Smart Sensing Business at Himax Technologies Inc., emphasized the need for user inference. “In today’s era of IoT,” says Chen, “the landscape is rapidly shifting towards the extreme edge, where everything must operate efficiently and with minimal power consumption.” To bridge this gap, Himax is committed to expanding its AI roadmap and introducing WiseEye2, an evolution of its successful predecessor, WE1.
WiseEye2 is specifically engineered as an ultralow-power AI processor tailored for on-device computer vision tasks. Its impressive features include 60mW facial mesh processing, a dedicated microNPU core, advanced power management, ample on-chip memory, and compatibility with an open software platform. This innovative processor caters to applications requiring immediate responsiveness and rapid decision-making, such as industrial quality inspection, predictive maintenance, and autonomous driving. It also serves well in applications prioritizing power efficiency and compact size, like wearables. Furthermore, its open software platform empowers developers to innovate and create new applications with ease.
Chen elaborates on the three significant enhancements that set WiseEye2 apart from its predecessor. First, the power architecture has been revamped, incorporating a multilayer power management mechanism and a multicore architecture to accommodate various use cases. This results in a remarkable 32 times improvement in inference speed and 15 to 50 times better inference efficiency compared to the first generation.
In terms of use cases, WiseEye1 focused on the person, face, or object detection, whereas WiseEye2 takes it a step further by not only detecting individuals but also discerning their actions and emotions by identifying facial landmarks. This level of precision allows for more accurate object analysis while maintaining low power consumption, marking a significant advancement in low-power AI computing.
Additionally, WiseEye2 boasts embedded industry-leading security and privacy features. The inclusion of Physical Unclonable Function (PUF) technology assigns a unique ID to each chip, which is utilized to create an encryption key for safeguarding software and ML features. Furthermore, WiseEye2 only transmits metadata results, not user image data, ensuring robust user privacy.
WiseEye2 also offers a developer-friendly toolchain and robust ecosystem support, utilizing Arm CPU and NPU framework to accommodate all available open-source AI models in the market. Chen provides real-world examples, such as doorbells and water meters, where WiseEye2’s ultralow power consumption plays a pivotal role in meeting customer demands for extended battery life.
Himax’s WiseEye2 garnered recognition in 2023 as the recipient of the Best AI Product award at the EE Awards Asia. While the AI landscape is evolving rapidly, Himax’s solution stands out as an ideal choice for resource-constrained, battery-powered endpoint AI applications, offering local inferencing tinyML vision AI capabilities, ultralow power consumption, a compact form factor, and robust security features.
Looking ahead to CES 2024, Himax is set to showcase its latest product developments and AI innovations, including AR/VR edge tracking solutions, advanced door lock technology, and automotive applications. The company’s vision of AI capabilities promises to revolutionize the smart cockpit experience in automobiles.
As we navigate an uncertain world, one thing remains clear: AI will continue to dominate the technological landscape in 2024 and beyond. Vendors and cloud service providers are intensifying their efforts to develop AI solutions tailored to various sectors. Himax is at the forefront, dedicated to advancing AI capabilities and seamlessly integrating AI into our daily lives, ultimately enhancing the human-machine interface and delivering an enhanced user experience.
Conclusion:
Himax Technologies’ WiseEye2, with its ultralow-power AI processing capabilities and focus on security and privacy, positions the company to thrive in the growing AI market. As AI continues to dominate the technological landscape, Himax’s dedication to innovation ensures a brighter future for the semiconductor industry, particularly in resource-constrained, battery-powered endpoint AI applications.
