All Categories
History
Price (excluding tax)
■Learning using only good product images, no annotation required DeepSky-SS (good product learning) is an AI (deep learning) system developed for production lines where there are no defective samples or where it is difficult to predict defects. It can be used in applications where it is necessary to detect defects that are difficult to define, such as when a line has just been started up and no defects have occurred, or when unexpected foreign objects have entered the line. All-in-one: "Image data collection → learning → judgment" can be performed with a single PC. No need to make detailed settings adjustments: just take pictures of good products and click the "Learn" button. AI automatically adjusts configuration parameters. Additionally, if a defect occurs, that image can be incorporated into the teacher image as a specific example of the defect. This improves the accuracy of detecting specific defects. ■What can be detected with DeepSky-SS ``I want to detect foreign objects, but the foreign objects vary.'' ``I don't know what kind of defective pattern will occur.'' - When there are no defective samples or a defective pattern cannot be assumed, DeepSky is a method that ``knows for sure that the product is not good.'' SS is best. Since good product learning uses only teacher images of non-defective products, there is no concept of ``types of defects.'' Therefore, for example, it is not possible to classify foreign objects into insects and vinyl, and defects into deformations and scratches. Instead, if the image contains information other than what is normally included in a non-defective product, it can be widely detected as a ``non-defective product.'' ■Differences from object recognition In object recognition, the defects you want to find and the things you want to count can be reliably learned through a process called annotation. Therefore, it is possible to detect small defects more reliably and to know the type, number, and location of detected objects. If you know the defect you want to find, it tends to be detected more reliably by having object recognition AI explicitly learn the defect. In this way, quality learning and object recognition have different roles, so it is recommended that they be selected or used together depending on the application.
You can search for other models from each index. The displayed value is the value of the currently selected part number.
CSV (Excel) data storage
Count number
I/O input/output
Image storage
Number of images used for learning
RS232C communication or socket communication
Remarks
Study time
Types of detection targets that can be registered per product
Part Number
DeepSky-SSProduct
AI visual inspection DeepSkyHandling Company
Skylogic Co., Ltd.Categories
| Image | Price (excluding tax) | CSV (Excel) data storage | I/O input/output | Image storage | Number of images used for learning | RS232C communication or socket communication | Study time |
|---|---|---|---|---|---|---|---|
|
|
Available upon quote | Divide and save by month (or day) by variety. | Optional intelligent I/O allows for type switching, inspection trigger input, and OK/NG/BUSY output. | Categorize and save in OK and NG folders for each type. (JPG, PNG) | Approximately 100 to 1,000 sheets | Switch types, start inspection, and output pass/fail judgment. It also returns the label name and number of detected objects. | 20 minutes to several hours |
Reviews shown here are reviews of companies.
Reviews shown here are reviews of companies.
