A recognition mode is a set of predefined rules that adjust the image recognition technology to operate efficiently under specific, commonly used conditions. Choosing the right recognition mode for your particular use case is crucial to ensure maximal technology performance.
Note that reference pictures pre-processing vary for different recognition modes. Thus, switching between recognition modes requires removing reference pictures and re-uploading them after the desired recognition mode is selected.
|Recognition mode||Use case||Color information||Availability|
|Single||A single object in a single query image.||Not used.||Configurable by the User.|
|Multi||A few or over a dozen of objects in a single query image.||Not used.||Configurable by the User.|
|Shelf||Several dozen or hundreds of objects in a single, large query image.||Used.||Contact us to activate.|
|Custom||You decide.||You decide.||Contact us.|
Designed to search for a single object in a single query image, assuming that the object size is similar to the query image's size.
It might return only a single matching object as well as a list of matching objects ordered by the confidence level. This means that, if it is desired, it is able to match several objects into a single region of the query image.
Designed to search for multiple instances of non-overlapping, potentially various objects in a single query image.
The result is a single matching object (the best one) for a respective region of the query image.
The Multi mode extension, optimized for shelf content recognition. Unlike the other modes, it uses color information to boost the matching precision.
Tell us what is the aim of your project and let us create a custom recognition mode for you.