Release of pixolution flow 2.1.1
This latest version of pixolution flow introduces new options to adapt the Keyword Suggester to your data set, along with better debugging possibilities when image loading fails.
Keyword Suggester can exclude images with too little keywords
For those who do not know our Keyword Suggester yet: It is a semi-automatic image tagging system. The keyword suggestions are calculated based on the keywords of indexed images that are visually similar to the input image.
The Keyword Suggester uses a statistical approach. Images with too little keywords will reduce suggestion quality since the variety of keywords is reduced.
To address this problem you can now set a threshold for the amount of keywords an image should at least have to be used for suggesting process. Images with no, one or only a few keywords are not suitable for the analysis. Now, those images can be excluded from the suggesting process.
Excluding images introduces a possible problem of not having enough data to suggest suitable keywords from. That's why you can now specify an overhead to internally request more images than actually needed for keyword suggestions. This overhead acts as a buffer to take images from when other images with too little keywords are ignored.
Easier debugging when loading of images fails
Loading images is an error prone task, especially when loading from the internet. To gain faster and greater insight why an image could not be loaded pixolution flow now returns all error causes within the response. You do not have to check the log file. This enables a faster and easier debugging process.
New product name pixolution flow consistently implemented
In summer this year we successfully redefined our product which is called pixolution flow since then. We are happy to say that pixolution flow is very popular with our customers. With the release 2.1.1 we finished the consistent naming scheme throughout our documentation as well as source codes.