Facebook Open-Sources Computer Vision
Facebook recently announced that it was opening its computer vision tools to the public. These tools are algorithms which help to identify, describe and label items in a photo. This revolutionary technology will change the future, from improving image searching on social media, creating experiences for Facebook’s visually impaired consumers to interpreting live videos as they unfold.
The Facebook Artificial Intelligence Research team (FAIR) has been working on computer vision for the past couple of years, in the hope to make it as good as actual human vision can be.
Facebook open-sourced three codes called DeepMask, SharpMask and MultiPathNet. The three different codes each have their own purpose, respectively searching for an object in an image (DeepMask), describing it (SharpMask) and finally identifying it (MultiPathNet).
There have already been huge advancements in the detection (finding an object) and classifying of images (labelling) thanks to deep neural networks that are constantly evolving and learning fresh patterns.
Facebook already has been focusing on a technique called ‘segmentation’ that makes use of algorithms to identify and make an outline of an object in a photo.
When fully developed, computer vision will have numerous possibilities in augmented reality. From identifying your food and telling you how many calories are in it, to virtually trying on clothes or new furniture at home, the opportunities are endless.
Facebook has released the technology in the hope that it will develop the tech a lot faster and new and exciting applications of it can be found. The social media giant has made the tech available for academics and researchers all over the world.
Once this progress is made, the next logical step will be to make the computer vision technology relevant to video as well. This is however a lot more difficult as in a video objects are constantly in motion.
Picture courtesy – aolcdn.com and bgr.in
by techtalks @TechTalks October 3, 2016 5:26 AM UTC