NEC develops AI for rapid 3D model generation from camera footage
NEC Corporation has developed new AI technology that can rapidly generate detailed 3D models solely from footage captured by general-purpose cameras, while automatically removing extraneous subjects.
Tokyo – NEC Corporation has developed an AI-powered technology capable of rapidly generating precise 3D models from footage captured by general-purpose cameras, such as those on smartphones. The technology automatically identifies and removes extraneous subjects like people or vehicles that could interfere with the modeling process.
The new method can produce high-quality 3D models in as little as one minute, significantly accelerating situational awareness in dynamic environments like power infrastructure facilities and construction sites. These models can be viewed on ordinary computers or tablets, enabling remote monitoring and rapid decision-making during anomalies or emergencies.
The technology combines NEC's proprietary AI with the Gaussian Splatting technique. It optimizes point density based on the visual complexity of the footage, reducing model generation time by up to 90% compared to conventional methods. Furthermore, the automatic removal of unnecessary subjects ensures that the generated 3D models accurately represent only the intended elements.
NEC Corporation aims to promote the adoption of digital twin technology within the infrastructure and construction industries. The goal is to address labor shortages and enhance worksite safety. The company plans to commercialize this technology by fiscal year 2027.