RUTILEA, Inc. (Headquarter: Kyoto, Japan; President: Takafumi Yano) announces the web version release of RUTILEA Defective Image Generative AI on Friday, August 18, 2023. The Defective Image Generative AI is a tool that allows users to easily generate a defective image from an image of a workpiece (metal, screw, food, capsule, tablet) they have on hand.
Example: generated defective pill image (the red arrow indicates the generated defect)
Automating inspection processes with AI requires defective image data as either training data or verifying data to ensure the accuracy of AI algorithms. Due to the low number of defective products, the lack of defective images can often become the bottleneck of an automation project besides the extra time and effort required to collect defective images from the past.
▶The correlation between the number of defective images and inspection accuracy
We conducted a verification case study regarding the correlation between the number of defective images and inspection accuracy. As a result, the accuracy of inspection by AI improved to 97% when using 900 generated defective images in algorithm training.
▶Easy to use Defective Image Generative AI
With RUTILEA Defective Image Generative AI, users can easily create defective images by selecting the defect type. The AI will then generate the defect of choice onto the original input image and create a new defective image. The free trial web version of Defective Image Generative AI is now available. Please feel free to give it a try.
▶Defective Image Generative AI web trial version (English available): https://generative-defect.rutilea.com/
▶How to use: please see here
▶Examples of generated defective images
▶Images of food contamination can also be generated.
For more examples, please see here.
▶RUTILEA provides problem-solving support for our customers to fine-tune the image generative AI catering to the customers' requirements. If you have a workpiece that is difficult to process in the web version or is of a rare material that makes it challenging to generate a defective image, please feel free to contact us.