The solution enables Aetina’s worldwide AI partners and clients to successfully implement edge AI through NVIDIA AI development and deployment tools and Aetina’s NVIDIA AI-powered training and inference platforms.
Jeremy Pan, AI Solutions Product Manager at Aetina, led the session and discussed a compelling use case of AI-powered automatic optical inspection (AOI) in a client’s plant in Yilan, Taiwan.
Challenges of Edge AI Adoption
System integrators and developers must train and deploy AI models on edge devices to implement edge AI. The AI model training process requires vast amounts of data to be collected and labelled utilising high-performance computing platforms, which can result in significant training costs. These factors make system integrators and developers’ training difficult and time-consuming.
AI model deployment might be tricky when system integrators and developers have several distant edge devices in different locations. AI model deployment concerns include optimising models for inference to operate quickly on edge devices, long deployment processes, difficulty with team communication and collaboration, high follow-up monitoring expenses due to many AI models, security risks, and expensive costs.
End-to-End AI Management Solution
Aetina offers a complete solution to assist its partners and clients in developing and deploying AI applications at the edge. Aetina’s NVIDIA-Certified edge computing platforms and NVIDIA’s AI model creation and deployment tools comprise the solution. Among these products are NVIDIA Fleet CommandTM and the NVIDIA AI Enterprise software suite, which provides enterprise support for the NVIDIA TAO toolkit and NVIDIA Triton Inference ServerTM.
Aetina’s NVIDIA-Certified edge computing platforms and other NVIDIA tools enable partners and clients to implement edge AI more quickly by streamlining the AI training and deployment process.
Ongoing application case in smart factory
One of Aetina’s clients, a global producer of industrial embedded flash and DRAM solutions, intended to improve efficiency by installing AI-powered AOI devices in its plant. Aetina’s client manufactures flash and DRAM products, which are small and complex electronic components designed for harsh environments and applications; the manufacturer of these components required an AOI system capable of processing high-resolution image recognition tasks at a high processing speed. Aetina worked with the customer to create a prototype of the AI-powered AOI system.
On Aetina’s AI training platform—SuperEdge AIP-D422—an Aetina solution team first trained an AI model capable of detecting defective goods in factory production lines for the AOI system with NVIDIA TAO, and then uploaded the model to NVIDIA GPU Cloud (NGC). The solution team remotely installed the model on Aetina’s AI inference platform—MegaEdge AIP-FQ47—in the factory of Aetina’s customer from NGC using NVIDIA Fleet CommandTM, successfully producing the AOI system prototype.
With AOI cameras, the system prototype can now substantially accelerate the inspection work of discovering DRAM goods with erroneously positioned corporate stickers with high precision in less than one second, compared to a human inspector who takes roughly 10 seconds to accomplish the same operation. When the AI-powered AOI system is finished, it will be put in Aetina’s client’s facility to execute inspection jobs across several production lines.
Furthermore, NVIDIA Fleet CommandTM enables the Aetina solution team to swiftly update the AI model on edge devices in Aetina’s client’s factory production lines, enabling effective remote AI management—a critical component of successful edge AI deployment.
The end-to-end AI management solution is a component of Aetina’s Pro-AI Service, which enables worldwide partners and clients to adopt AI for vertical applications other than AOI in factories, using Aetina’s edge AI hardware and software.