Our models are designed for travel and trained on travel images only. Therefore the accuracy of this models is very precise. But beside the accuracy our approach is “semantic concepts”, to imitate human behavior when someone is looking for travel.

No one is looking for just tags like “chair” or “cloud” or “cocktail” In travel people look for experiences they can relate to like “romantic dinner on the beach on a Thai island” or “Rooftop pool in a 4 star hotel in Bangkok” or ” Romantic beach hut on Bali in November” Those are searches we can support with our algorithms.

“Experience & Activity” tags (API). These are specific trained tags that occur images on Online Travel Agencies and meta search sites. See the complete list below.

These are the semantic concepts we were talking about here above. Real travel searches translated in a sophisticated image detection model. “Beach huts”, Romantic dinner on the beach”, “Riverside walks” and much more. For the complete list of tags contact us.

Conversion image quality (API)

This classifier is as simple as it is sophisticated. We have trained this algorithm on more then hundred thousand images. For you available to tag images with an aesthetic score between 1 and 100%. Great for filtering.

The output in a percentage of the perfect image lets you easily set a threshold to ban bad images from your site.

General Travel image tags (API)

This classifier is trained to recognize over 4000 thousand concepts linked to travel. We selected those 4000 concepts out of a model that has even much more concepts. The travel image tagging classifier gives confidence scores about what could be in the image. Per image you get 5 items.

This classifier is used to categorize your images by whats in it. You can extract your own concepts from that by combining our concepts. For example. There is beach and golf court in your set of images. You can target specific client groups on activity or “experience”

Click here to see all labels from all classifiers to get inspired about your new landing pages and content filtering.

0 Shares