Researchers in Japan have introduced LatentCSI, a technology that uses artificial intelligence to generate detailed images of a room's interior based on an analysis of Wi-Fi signals. An ordinary router can now gain an unexpected ability: the power to "see" a room's layout, thanks to a new method developed by a team from the Tokyo University of Science.
The technology is based on analyzing so-called Channel State Information (CSI). Wi-Fi waves constantly bounce off walls, furniture, and people, creating a unique map of a space. While this data previously only allowed for rough schematics to be created, that has now changed.
The new approach, named LatentCSI, uses an AI model similar to those in image generators, but with a key difference. Instead of converting raw radio data directly into pixels, the system translates it into a compressed digital representation, which serves as the foundation for building a detailed picture. The technology doesn't generate an image from scratch; it is first trained on real photographs of the specific room. After training, the AI uses live Wi-Fi signals to reconstruct the current scene: the arrangement of furniture, the number of people, and their locations.
A significant advantage of this method is its efficiency. Operating in this "latent space" requires less computational power compared to previous Wi-Fi visualization attempts, making the process faster. Despite these impressive results, the technology has a major limitation: it cannot create an image of an unfamiliar room. To work, it requires reference data from being pre-trained on that room's photographs.
The development of such systems unlocks the potential for new surveillance tools, where a common router could become part of a monitoring system. For now, however, this development remains a laboratory project, showcasing new possibilities at the intersection of wireless communication and artificial intelligence.
