AOI inspection machine VS AI visual machine ?

Both have advantages and disadvantages.
In terms of development speed, the development speed of AI can be faster purely for the algorithm part.
Judging from the detection speed, it is not necessarily possible to tell which one is faster. It depends on the content of the detection and the complexity of the background.

• The product needs to have enough defective samples so that it can learn quickly. High-performance AI can learn 100 PCS 5M NG photos in about 2 minutes. And can generate source code and add it to the system program.
• If the product needs to adjust the algorithm due to process changes, it can also learn quickly and generate empirical parameters to add to the system program.
• Products need to accumulate experience to adjust the algorithm. Each batch of products uses NG or misjudged parts for rapid learning, and can generate experience parameters and add them to the system program.

• 當瑕疵檢測需要極高速的運作如每秒100PCS以上，這種高度優化以MCU為運算主軸時，傳統程式可以調整的空間大，有較佳的表現。
• 當瑕疵檢數量極少，產品良率都在99.7%以上時，NG的樣品取得不易，AI的方法很難處理未知的瑕疵，傳統的程式 卻可以以一般平面或是邊緣檢測的概念處理得很好。
• 當需要少量多樣時，無法經常停機做深度學習時，傳統的程式 可以以一般通面或是邊緣線做檢測，可以適應的效果較佳。