Over the past few years, Facebook has been improving its image recognition capabilities, which will soon give marketers a whole new consideration. Today, Facebook released a new research paper outlining their latest image recognition developments, including advanced object-based search capabilities.This is how Facebook image search is going and the potential opportunities it will provide. computer vision Last April, Facebook announced the release of automatic alt text or automatic alt text for images posted to Facebook. Automatic alt text uses object recognition technology to generate photo descriptions, which are processed by Facebook’s artificial intelligence engine to build image content.
The system is the culmination of years of work – back in November 2015, Facebook showed progress with its image recognition AI, with a system capable of distinguishing objects in photos 30% faster and using less training data , higher than previous industry benchmarks. Facebook’s Evolving Image Recognition Technology Offers New Opportunities UAE Phone Number List Social Media Today After years of work, the level of detail the system now translates doesn’t seem to pay off much, but the amount of work involved in teaching a computer to “see” an image is enormous. The system needs to be trained on huge datasets, with associated descriptions of each image, in order to identify key elements and determine what each object is.
However, even with all this training, there are still limitations to what image recognition can do – for example, in the image below you can see that when objects “overlap” each other, the system still stumbles, recognizing two people as one . Facebook’s Evolving Image Recognition Technology Offers New Opportunities Social Media Today These types of more limited recognition abilities are difficult to train, which gives you some idea of the challenges Facebook researchers faced when building an image recognition system. For systems based on code and input, things that humans can easily do are not simple.
Again, it doesn’t look like it’s breaking ground, and probably won’t have a big impact on users who aren’t visually impaired, but it’s developing a lot. The fact that Facebook’s systems are now taking action underscores how smart their systems are getting, and the amount of work involved in getting to this stage is staggering, although Facebook does have a massive training model at its disposal, with daily Nearly 2 billion photos were uploaded to Facebook, Instagram, Messenger and