Just as we develop a feeling for a “good image” through the many hundreds of thousands of images that we transport to the digital world for our customers in the course of digitization projects, most people who look at and evaluate images on a daily basis will probably be able to do so. We see whether an image is sharp and well exposed, whether the tonal values are untrimmed and whether it makes a coherent overall impression.
Can this assessment of the quality of an image be supported or even replaced by machine learning systems? How do you assess the beauty of a landscape or portrait, for example, to evaluate images?
So what is a good picture? You can answer this difficult and partially subjective question with the latest release of our Clarifai integration “AutoTagging for Cumulus” – at least when it comes to evaluating the technical quality of an image.
The rating is done by using the following tagging models:
You can try it yourself. “AutoTagging for Cumulus” – our solution for supporting the capture of metadata through machine learning.