株式会社Roxy Company Profile

The high-speed, high-precision appearance inspection AI 'Roxy AI' has been upgraded with a focus on the usability of the runtime environment. ■ Easily Define Inspection Flow Roxy AI possesses the flexibility to accommodate various inspection patterns. However, this sometimes resulted in longer initial setup times. In version 2.1.0, we have templated representative inspection sequences. By simply selecting the desired inspection sequence and adding the minimum necessary settings, you can start operations immediately. ■ Scripts for Flexible Inspections The script feature significantly enhances inspection flexibility. However, since programming is required, it can be a barrier for those who are not familiar with it. In version 2.1.0, we have included practical sample scripts for rearranging defect details, integrating defect types, and pre-checking changes in the shooting environment. By modifying these as a base, it becomes easy to meet specific customer requests. ■ Others We have strengthened operational usability in various ways, including improving inspection speed by several tens of percent and enabling inspections with AI trained in the V2 environment to be conducted in the V1 runtime.

High-precision, high-speed appearance inspection AI 'Roxy AI' has undergone its first major version upgrade! With this addition of features, we have lowered the barriers to implementing AI inspections more than ever before, achieving stabilization of inspection process quality and cost reduction. ■ Attention Map A representative method for defect detection is semantic segmentation (hereafter referred to as semseg). High-performance models such as Yolo have been released, and many appearance inspection AI products have adopted them; however, the annotation for semseg is very cumbersome, which has become a bottleneck for implementation. When the boundary between normal and defective is ambiguous, it is even difficult to annotate correctly. With the newly added unique 'Attention Map' in Roxy AI V2, this problem is dramatically improved. There is no need to be conscious of the boundaries between normal and defective; simply clicking on the defective area is sufficient. This achieves semantic segmentation in an overwhelmingly efficient manner. ■ Learning Environment - Significantly improved detection capability of small objects in measurement AI - Official support for RTX 50 series - Support for learning on CPU - Other numerous improvements ■ Runtime Environment - Compatible with KEYENCE's LJ-S8000 - Added a group of workers to flexibly control inspection results

The high-precision, high-speed appearance inspection AI 'Roxy AI' has been upgraded. With this addition of features, we have lowered the barriers to implementing AI inspections more than ever before, achieving stabilization of inspection process quality and cost reduction. ■ Learning Environment - Added model stability check function - You can understand the stability for each label - You can pre-check resistance to changes in imaging conditions - Added support for measurement AI that can be used without learning - Enhanced alignment function - Enhanced data cleaning function ■ Runtime Environment - Supports ModbusTCP through PLC integration ■ Common to Learning and Runtime - Supports USB dongle licensing