Google Rolls Out New Version of “Open Images” with Localized Narratives

Back in 2016, Google took a step towards building improved artificial intelligence that would be able to recognize a lot more objects than might have been the case at that specific point in time. Most people agreed that AI was the way of the future, but a lot of these people really struggled to find a way to make it so that the AI that was being developed could truly be advanced.

Hence, Google took a step that was to create a dataset that they referred to as “Open Images”, and this dataset contained nine million photos with various labels attached to them. This was to help AI learn how to judge certain situations, with the goal being to help it achieve some amount of sentience at least in a small way.

This project has been quite successful and has led to a fair amount of advancement in the field of AI research, but at the same time Google is known for continually expanding its efforts and this is why the tech giant has rolled out a new version of Open Images, one that contains even more photos as well as labels.

However, more than the twenty three million new high quality labels that were created for AI to use, this new version of open images contains localized narratives which basically means that users can hover over certain parts of images in order to describe what is occurring. This will help AI get used to certain colloquialisms as well as gain a deeper understanding of the complexities of every day objects. Only time will tell what kind of benefits this will have for AI that people might just end up using, but it’s fair to say that these developments make it clear that we are living in very exciting times.




Read next: Google says its AI will tag man and woman in photos as person to avoid unfair bias

No comments:

Post a Comment