Researchers at the Institute of Science, Tokyo, developed LatentCSI, a system that converts Wi-Fi signal reflections into photorealistic images of indoor spaces. The technology maps Channel State Information into latent space and uses a modified version of Stable Diffusion 3 to generate detailed room reconstructions. The artificial intelligence system requires initial training with photographs of a specific location before it can interpret live wireless data to show occupant positions and spatial layouts.
The method cannot function in untrained environments, but it demonstrates advances in wireless sensing capabilities that surpass previous blurry imaging attempts. Motion-sensing features are already present in some consumer routers, and future applications of similar technology could enable home automation systems or raise surveillance concerns. The research project demonstrates how wireless signals, combined with generative models, can produce visual information without the use of cameras.
The method cannot function in untrained environments, but it demonstrates advances in wireless sensing capabilities that surpass previous blurry imaging attempts. Motion-sensing features are already present in some consumer routers, and future applications of similar technology could enable home automation systems or raise surveillance concerns. The research project demonstrates how wireless signals, combined with generative models, can produce visual information without the use of cameras.
