Want to differentiate your industrial equipment? AI in manufacturing serves as a primary source of competitive advantage, enhancing productivity in production lines and warehouses. Metis® – Axelera® AI’s AI acceleration platform – helps you adopt and accelerate edge AI in manufacturing. With Metis, your machine vision and Industry 4.0 equipment can deliver unmatched AI performance.
In Industry 4.0, computer vision, machine vision and robot vision are applied for e.g. product assembly, defect detection, quality control, workplace safety inspection. High resolution and high frame rate are essential for these applications. While moving data to the Cloud is effective for quick proof-of-concepts, it increases latency limiting any real-time decision, and raises privacy and security concerns by transferring sensitive data away from the owner's control and exposing it to external threats. With industrial edge solutions, you can carry out all your Al inferencing locally .
For all industrial edge applications, industrial-grade reliability, low power consumption for extended temperature range and fanless design, as well as low latency and high throughput are the common challenges in AI hardware. Axelera AI’s Metis platform answers these challenges. Metis offers industrial equipment manufacturers vision AI performance that other accelerators cannot achieve. Metis even enables industrial AI acceleration at a GPU level where power-hungry GPUs are not a viable option.
With its unbeatable combination of high performance with low power consumption, the Metis Al platform is the best choice for AI at the edge. The Metis AIPU is a game-changer, as it can perform up to 214 trillion operations per second at ultra-low latency without sacrificing accuracy. It enables manufacturers and machine builders to enhance productivity and customer satisfaction developing energy-efficient and cost-effective applications at the edge.
AI-based visual inspection systems utilize machine learning algorithms to analyze and evaluate product images in real time. The applications are endless, including use cases which were simply not possible before, like real-time sorting of garbage or vegetables with a level of granularity previously unattainable without neural networks. Furthermore, these systems enable faster and more flexible Optical Character Recognition (OCR), advanced defect analysis, precise pick-and-place operations, and sophisticated 3D vision in robotic applications. Innovation on the production line is all about AI.
Time-motion study has been a business efficiency technique since the early 20th century. Hundred years later, video systems can provide extremely accurate insights for boosting workers productivity while protecting their privacy. The same systems can also monitor occupancy storage and traffic areas to optimize picking and packing. Autonomous Mobile Robots are redefining the way goods are moving inside a warehouse, requiring a navigation system based on 6 or more cameras to ensure reliable and safe operation. Similarly, in vertical automated warehouses, computer vision AI can help operators reduce error rates by alerting employees handpicking wrong products from the drawer, thereby improving accuracy and reducing wastage.
Machine builders are leveraging AI to create value for their customers and get insights on the way the machine is operating. Integrated video analytics can detect defects in real-time, improving customer satisfaction through better quality and increased productivity. Simultaneously, these systems can identify patterns in machine outputs that signal the need for maintenance before a failure occurs. Another critical area is worker safety. In 2021, the manufacturing industry had one of the highest rates of nonfatal work injuries and illnesses, surpassed only by healthcare and retail. Video analytics can enhance safety by slowing down machine speed when an operator approaches and stopping the machine if biometric data indicate that the worker is unauthorized.
The highly energy-efficient, easy-to-integrate and future-proof Axelera AI Metis Platform accelerates equipment and software deployment by providing a comprehensive hardware and software solution for accelerating Vision AI in Industrial Manufacturing. While the hardware design enhances performance or reduces cost and power consumption, the usability of Metis makes vision AI technology accessible to a broader range of developers and enterprises worldwide, so they have the freedom to innovate.
The Metis AIPU stands out by offering this transformative capability with industry leading power and cost efficiency. Voyager® SDK (link), our easy-to-use and flexible toolchain built upon open-source components, supports object detection networks like YoloV7 and YoloV8, segmentation networks like UNet, different OCR networks, and many more. It also provides full flexibility to optimize performance and customize the computer vision pipeline for the specific application.
The Metis AI platform ensures that AI in manufacturing meets the rigorous demands of industrial applications, and does so with unparalleled cost efficiency and operational reliability, all while keeping data secure onsite.
Our cutting-edge chip features a power envelope of less than 15W, ensuring efficient energy usage while delivering exceptional performance. With the capability to downclock, you can fine-tune the power consumption to meet your specific needs, all while maintaining an outstanding frames-per-second per watt (fps/W) ratio. This balance of power efficiency and performance makes our chip the ideal choice for high-demand applications.
Our networks are automatically quantized during compilation to run efficiently and accurately on Metis. The accuracy after quantization remains comparable to the original floating-point networks. For example, two popular networks, Resnet-50 and Yolov7, each achieve quantized accuracy within 0.17% of the original 32-bit floating point models.
The Metis AIPU offers a unique blend of cost-effective and energy-efficient AI inference capabilities. Moreover, a single chip can deliver up to 214 trillion operations per second. Our PCIe card powered by 4 Metis AIPU achieves record-breaking performance and efficiency.
Getting started with the Metis AI Platform is easy thanks to our end-to-end evaluation systems, based on both x86 and ARM platforms. Experience the inference and accuracy capabilities of our Metis AIPU, and work on integrating your application for some real-world benchmark. Integrating the Metis AIPU into your products will then be simplicity itself, thanks to our acceleration cards based on M.2 and PCIe standard form factors .
What’s more, our low-code Voyager SDK speeds up software development and system integration, helping you bring your industrial equipment to market fast.
Tweak one of our pre-trained AI models or develop your own – with our unique quantization method you may only need a short calibration based on a small subset of your existing dataset. Either way, our integrated systen ensures optimized end-to-end performance and the highest accuracy possible. Supporting open standards including OpenCL and SYCL, the Voyager SDK integrates seamlessly into your existing development flow for maximum ease of use.
Industrial Manufacturing applications come in all shapes and sizes. However, the Metis AIPU, the most scalable AI accelerator technology on the market is available, so you can easily support all your projects with a single platform. We offer single-AIPU M.2 acceleration cards and a PCIe acceleration card, letting you balance cost and performance. Need more AI power for larger optical inspection lines or high efficiency for your embedded system? Axelera AI’s portfolio can cover all your different needs with a scalable range of products and a strong ecosystem of partner solutions.
Axelera AI today announced Metis M.2 Max, a new addition to the company's class-leading Metis AI processor unit (AIPU) family. The Metis M.2 Max delivers first of its kind performance for LLMs and VLMs for low power, embedded devices.
Axelera AI, today announced the expansion of its leadership team with three key executive additions: Marta Ostroumoff joins the company as Chief Financial Officer, Eke Bijzitter has been appointed Head of Legal, while current CMO Alexis Crowell will expand her scope adding General Manager of the Americas to her responsibilities.
Axelera AI, the leading provider of purpose-built AI hardware acceleration technology for generative AI and computer vision inference at the edge, today launched the Axelera Partner Accelerator Network, a global partner program designed to accelerate the development of customer-ready solutions at the edge using Axelera technology.
