Searching for images on the Internet can be hit or miss. That’s because most image searches rely on metadata (text associated with the images, such as file names or dates), and metadata can be incomplete–if it’s there at all. Software that analyzes the images themselves has been notoriously unreliable. But it could get a boost from a technology developed at the University of California, San Diego.
The technology is based on existing systems that learn to describe pictured objects in terms of features like color, texture, and lines by practicing on pictures in a database of known objects. The UCSD system adds a new twist: it assigns each image a likelihood of belonging to categories such as “sky,” “mountain,” or “people.” Then it uses those words to label parts of the pictures. The technique is 40 percent more accurate than typical content-based image-search methods, says Nuno Vasconcelos, a UCSD professor.
Here’s how a Twitter engineer says it will break in the coming weeks
One insider says the company’s current staffing isn’t able to sustain the platform.
Technology that lets us “speak” to our dead relatives has arrived. Are we ready?
Digital clones of the people we love could forever change how we grieve.
How to befriend a crow
I watched a bunch of crows on TikTok and now I'm trying to connect with some local birds.
Starlink signals can be reverse-engineered to work like GPS—whether SpaceX likes it or not
Elon said no thanks to using his mega-constellation for navigation. Researchers went ahead anyway.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.