Google Releases TensorFlow 3D to Boost AR Apps Using LiDAR Sensors
Google introduces TensorFlow 3D, a new library that helps developers build advanced augmented reality experiences using LiDAR and depth sensors. The toolkit provides essential models for 3D scene understanding on mobile devices.
Google introduces TensorFlow 3D, a new library that gives developers tools to leverage LiDAR and depth sensors for advanced mobile augmented reality experiences. This release brings a resurgence of depth-sensing technology, which gained prominence through Apple's LiDAR-equipped iPhones and iPads, by providing a structured way to process complex 3D data.
The new toolkit includes 3D semantic segmentation, 3D object detection, and 3D instance segmentation models that enable applications to understand and interact with physical environments. For example, the 3D semantic segmentation model allows apps to distinguish between foreground objects and backgrounds, similar to the virtual background effects popular in video conferencing apps.
According to Google researchers, entering the field of 3D scene understanding is often difficult due to a lack of accessible resources, and TF 3D solves this by providing operations, loss functions, and data processing tools. While the library supports autonomous cars and robots, its mobile AR capabilities enable developers to create highly accurate spatial mapping and interactive experiences on modern smartphones.