Facebook Image Search And Auto-Tagging Technology: Search An Image By Action, Place Or Color
In a continuous struggle to make Facebook user-friendly and reliable, Facebook has introduced a new feature of tagging images automatically. Auto-labeled images will pop-up in the search results. Images once auto-tagged will not require text or keywords.
According to Joaquin Quiñonero Candela, the director of machine learning at Facebook, that Facebook is building scalable systems to make content easy for both machine and human reading. Joaquin said that Facebook is revolutionizing search-engine approaches by introducing pure image search. It's really a novel and innovative concept that will open new horizons in social media, e-commerce and online job market.
Current search engines operate under a text driven technology which is purely text dependent and cannot read images. New technology will understand images at the pixel level progressing to the next stage in the artificial intelligence and search engine arena.
Engineering a system that can understand images and things visible in the image is really a task beyond average intelligence. The new innovation doesn't seem limited to image searching but it can be applied to public data, criminal records, research purposes, investigations, Education much more.
Into the Facebook, an image once shared will be tagged automatically by computer. Even without text description or caption, the Facebook image recognition feature will find it out. This technology has been experimentally tested for finding and removing nude content.
According to Digital Trends, this technology will enable searchers to find a place, an action, a particular garment, color based objects and even faces.
This technology can be used to enable future Robots to recognize human beings, colors, trees, hills, water, roads and other things of daily use through combination of artificial intelligence. Through Lumos Algorithm, this technology will automatically update itself with every bit, every feed and every update with an image or a text.
Facebook machine learning teach trained the system with 130,000 actual Facebook photos and different actions. According to candela, the idea is extendible to video recognition.