• Adacotech Incorporated

In many corporations, the adoption of exterior appearance inspection AI is limited to trials that do not translate to its use in mass production. One of the main reasons for this is failing to build the envisioned models due to the use of insufficiently processed image data in training, resulting in unstable inspection accuracy. This technology provides a mechanism that relies on visualizing the selection and feature distribution of the training data to accurately ascertain the quality of the data when building an AI model. In addition, high speed training makes it possible to achieve shorter model refinement cycles. Therefore, even when the amount of data limited, a stable exterior appearance inspection AI model usable on mass production lines can be built.