Facebook’s AI has X-ray vision and sees you over Wi-Fi

DensePose can estimate body shape from the decoded signals. (photo: clone)

🔊 Listen to this news by clicking here

When George Orwell imagined the world of 1984, cameras and eavesdropping devices were the protagonists of the surveillance that the work proposed. Now, new ways of “monitoring” in certain spaces may be breaking the barrier of more modern technologies, such as Wi-Fi. A study conducted by the artificial intelligence (AI) division of Meta, the holding company of Facebook, identified a way to find out exactly where and what movements people make over an internet connection.

The research, called DensePose, claims that an array of a Wi-Fi signal with machine learning technology applied to the architecture can “scan” up to 24 regions of the body to produce the image corresponding to the individual being monitored.

According to the Meta, the tool is “Facebook’s real-time approach to map all human pixels from 2D images into a 3D model based on the surface of an object.” The technology has caught the attention of researchers outside Meta, who are also studying the project.

DensePose is open code that converts Wi-Fi signals reflected by an object in the environment into coordinates, creating a kind of silhouette image that is decoded on a computer.

“We developed a deep neural network that maps the phase and amplitude of Wi-Fi signals to UV coordinates in 24 human regions,” says the study. “The results of the study reveal that our model can estimate mode intensity for multiple people, with similar visual performance to image-based approaches, using Wi-Fi signals as the only input. This paves the way for low-cost, widely accessible, and privacy-preserving algorithms for human detection.” .

According to the researchers, this would be a cheaper way to develop spatial scanning-based equipment, such as lidar, that processes image depth.

The AI ​​R&D market has grown over the past few years and is getting closer to the users, as tools like ChatGPT and Lensa, the image creator app, reach the consumer. However, the level of information captured by these AI systems is one of interest to users and researchers, since many of them are associated with invasive use of data, images, and even surveillance.

For DensePose, there are still no projections for widespread use or commercialization of the technology, but studies have been developing within Meta since 2018.

Counting
All technology