AI Visual Inspection Product "Roxy AI"

The AI used for visual inspection places the utmost importance on accuracy. Many visual inspection AI products aim to improve accuracy by adding data or repeatedly tuning parameters until accuracy increases. To achieve high precision under the condition of having a limited number of defective samples, Roxy AI emphasizes data organization. In this version upgrade, various features have been added to make it even more user-friendly and to achieve even higher accuracy. - Judging the quality of shooting conditions before AI learning: It is often discovered after trial and error that the reason for low AI accuracy was poor imaging conditions. Roxy AI detects adverse conditions that could negatively impact AI learning before the learning process begins, minimizing unnecessary trial and error and rework. - Automatic data organization: There is no need to manually check large amounts of data thoroughly. By automatically excluding data deemed inappropriate by the AI during training, the time and labor required for data organization are reduced. - Additionally, an integrated viewer that allows observation of data from various perspectives, among other features.

We have released Roxy AI V1.12. Various features have been added to achieve further improvements in accuracy and usability. - Expansion of data organization methods The organization of training data, a distinctive feature of Roxy AI, has been significantly enhanced! By organizing data from various perspectives throughout the entire learning process, we achieve higher precision in AI creation. - Addition of new AI models Compared to previous AI models, accuracy has been further improved! You can select the appropriate model according to the characteristics of the inspection target. - Visualization of contributing factors to judgments A feature has been added to visualize where the AI determined defects! You can visually confirm that the AI is accurately identifying defects. Additionally, it is possible to visualize how the AI captures images at each layer. This can be utilized when fine-tuning accuracy. - Furthermore, numerous features have been added that contribute to improved accuracy and usability.

We have released Roxy AI V1.11. Various features have been added to achieve further improvements in accuracy and usability. - Added image integration feature (patent pending) The image integration feature is very effective when using KEYENCE's LumiTrax to capture and inspect multiple types of images (such as shape images). It allows for annotation, learning, and inspection of multiple types of images at once, improving accuracy, shortening inspection time, and reducing the effort required for AI creation. - Added new AI model Compared to previous AI models, accuracy has been further improved. The inspection speed remains high, similar to previous AI models. - Enhanced detection capability for unknown defects Roxy AI can detect not only learned defects but also unknown defects. In this version, the detection capability for unknown defects has significantly improved. - Added variations for data organization Data organization is a feature of Roxy AI. In this version, we have enhanced functions to extract and organize inappropriate data during learning, as well as boost functions to strengthen weaknesses. - Additionally, many features have been added that contribute to improved accuracy and usability.

We will be exhibiting at the "Nagoya Smart Factory EXPO 2021" held at Port Messe Nagoya from October 27 (Wednesday) to 29 (Friday). In addition to a demo of Roxy AI's model creation, we will introduce a runtime that allows for the easy construction of inspection systems enhanced in version 1.10.0, as well as a demo of AI inspection using collaborative robots. We look forward to your visit.