Defect feature values, set via neural network rather than manually

SUALAB, an AI-based smart factory solution provider, has officially launched 'SuaKIT,' a deep learning-based machine vision inspection software.

Based on image interpretation technology, Suakit has significantly enhanced the accuracy and speed of inspections for various areas that were difficult to inspect with existing machine vision technology.

Previously, engineers had to define defects one by one on images and manually set their characteristic values. Consequently, machine vision technology was applicable only in fields where surface shapes are standardized, such as semiconductors or LCDs. Conversely, in fields where surface shapes are irregular, such as textiles or natural leather, it was difficult to manually set characteristic values, so visual inspection was mostly relied upon.

By collecting a small number of images of normal and defective products using Suakit's deep learning technology and training a deep learning algorithm, the artificial intelligence neural network automatically identifies defect feature values. Since there is no need to manually set defect feature values, it can be utilized in manufacturing fields such as textiles and leather where surface shapes are irregular.

In addition, it is designed to process data at high speeds on high-performance GPUs through CUDA (Compute Unified Device Architecture) technology. This enables Suakit to perform at its best even in manufacturing processes that require high speed.

Suah Lab CEO Song Ki-young expressed confidence, stating, “As Suah Lab’s current deep learning-based machine vision technology is unrivaled, we will build entry barriers even faster in the future and become an unrivaled company in the factory automation sector.”