Composite Finder is a system that extracts arbitrary situation images from a group of images.
When the user inputs the image to be extracted from the image group to be registered in the database, the situation image similar to the input image will be picked up.
Images can be extracted using the following two patterns in Composite Finder.
- Image Search: Extracts similar situation images in the entire image
- Image range search: Clips a part of the image and extracts the image containing the range similar to the situation in that area.
Composite Finder does not require object labeling when registering new images in the database.
Usually, the learning data / verification data for developing various cognitive / judgment applications is often picked up by the user by human confirmation, which requires a huge amount of work.
In addition, there is concern about the problems such as work errors for human work, and it can be in a factor that it is not possible to visualize judgments as numerical values, which may lead to personalization work, and in factors that hinder traceability.
Situation extraction in special situations (example: factory, in the plant) has few products that have entered the market, enormous work is required even if it is made in-house, and a group of images taken of the factory even if it is outsourced. There are many cases where it is confidentially difficult to deploy to an outside company.
By introducing Composite Finder, it is possible to automate the work, and it will be possible to solve the above problems.
- Extract situation trends using image analysis technology when building a database
- The composition of the scene to be extracted specified by the user is used as input, and an image with a tendency similar to that image is extracted from the image group.
- The composition of the scene to be extracted specified by the user is used as input, a part of the image is clipped from the image group, and the image having a tendency similar to the situation in that area is extracted.
Use Case Example
Development of various recognition / judgment applications
Since similar images are automatically extracted, a significant reduction in development man-hours can be expected.
Automatic creation of teacher data required for deep learning / machine learning models (automatic annotation)
By linking with the annotation tool, the man-hours for annotation that are currently performed manually can be significantly reduced.
When building a database, by storing images + various sensor information in the database as a set, it is possible to analyze what factors occurred before and after the same situation.
Example) Comprehensive analysis of what kind of situation was when a false detection of the sensor occurred
|License Type||Initial license *1|
|Maintenance service *2|
|Provided Unit||1 terminal / 1 license *3|
|Offer Form||Web download *4*5|
|License Price||Please contact the following separately.
Inquiries: To the product window
|Sales Start Date||February 1, 2021|
*1 Includes maintenance service fee for the first year.
*2 Maintenance service must be renewed once a year.
*3 As a system configuration, the number of installations and the number of DB instance accesses are linked.
*4 A USB dongle is available as an option when requesting provision on a device. Please contact us when making a quote.
*5 As an option, it is possible to expand the access function from a web browser. Please contact us when making a quote.
We provide a catalog of our products in PDF format.
If you wish, please complete the questionnaire below and download it.
* The screen is under development. It may be different from the product version.