17592500. CAMERA LOCALIZATION simplified abstract (MICROSOFT TECHNOLOGY LICENSING, LLC)
Contents
CAMERA LOCALIZATION
Organization Name
MICROSOFT TECHNOLOGY LICENSING, LLC
Inventor(s)
Sudipta Narayan Sinha of Kirkland WA (US)
Joseph Michael Degol of Seattle WA (US)
Tien Do of Minneapolis MN (US)
CAMERA LOCALIZATION - A simplified explanation of the abstract
This abstract first appeared for US patent application 17592500 titled 'CAMERA LOCALIZATION
Simplified Explanation
The patent application describes a method for camera localization within a scene using machine learning. Here are the key points:
- The method involves inputting an image of a scene captured by a camera into a machine learning model.
- The machine learning model has been trained to detect a variety of 3D scene landmarks that are pre-specified in a pre-built map of the scene.
- The machine learning model outputs predictions, which can be either a 2D location in the image representing one of the 3D scene landmarks or a 3D bearing vector pointing towards the predicted 3D location of a scene landmark.
- Using these predictions, the method computes an estimate of the camera's position and orientation within the pre-built map of the scene.
Potential applications of this technology:
- Augmented reality: This method can be used to accurately position virtual objects within a real-world scene, enhancing the user's augmented reality experience.
- Navigation and mapping: By accurately localizing the camera within a scene, this method can be used for navigation purposes or to create detailed maps of indoor or outdoor environments.
- Robotics: This technology can be applied to robotic systems to enable them to understand and navigate their surroundings more effectively.
Problems solved by this technology:
- Accurate camera localization: The method provides a reliable way to estimate the camera's position and orientation within a scene, even in complex or changing environments.
- Scene understanding: By detecting and localizing 3D scene landmarks, the method improves the camera's understanding of the scene, enabling more advanced applications such as augmented reality or robotic navigation.
Benefits of this technology:
- Improved accuracy: The use of machine learning allows for more precise localization of the camera within the scene, leading to better performance in applications like augmented reality or navigation.
- Flexibility: The method can be trained for different scenes, making it adaptable to various environments and scenarios.
- Real-time performance: The method can provide camera localization estimates in real-time, allowing for immediate feedback and interaction in applications like augmented reality or robotics.
Original Abstract Submitted
In various embodiments there is a method for camera localization within a scene. An image of a scene captured by the camera is input to a machine learning model, which has been trained for the particular scene to detect a plurality of 3D scene landmarks. The 3D scene landmarks are pre-specified in a pre-built map of the scene. The machine learning model outputs a plurality of predictions, each prediction comprising: either a 2D location in the image which is predicted to depict one of the 3D scene landmarks, or a 3D bearing vector, being a vector originating at the camera and pointing towards a predicted 3D location of one of the 3D scene landmarks. Using the predictions, an estimate of a position and orientation of the camera in the pre-built map of the scene is computed.
