18243422. PROBABILISTIC APPROACH TO UNIFYING REPRESENTATIONS FOR ROBOTIC MAPPING simplified abstract (Samsung Electronics Co., Ltd.)
Contents
PROBABILISTIC APPROACH TO UNIFYING REPRESENTATIONS FOR ROBOTIC MAPPING
Organization Name
Inventor(s)
Jungseok Hong of Minneapolis MN (US)
Suveer Garg of New York NY (US)
Ibrahim Volkan Isler of Saint Paul MN (US)
PROBABILISTIC APPROACH TO UNIFYING REPRESENTATIONS FOR ROBOTIC MAPPING - A simplified explanation of the abstract
This abstract first appeared for US patent application 18243422 titled 'PROBABILISTIC APPROACH TO UNIFYING REPRESENTATIONS FOR ROBOTIC MAPPING
Simplified Explanation
The present disclosure provides methods, apparatuses, and computer-readable mediums for evaluating a reliability of three-dimensional shape predictions. In some embodiments, a method includes obtaining a scene representation including one or more images, estimating semantic information of partially-observed objects in the scene representation, based on geometric information extracted from the one or more images, determining segmented point clouds of the partially-observed objects based on the semantic information and the geometric information, creating metric embeddings of the segmented point clouds corresponding to the partially-observed objects, predicting completed shape point clouds of the partially-observed objects, generating, using a probabilistic machine learning model, confidence scores for each point of the completed shape point clouds, based on a correlation between the geometric information and the semantic information, and controlling a motion of a robot based on the confidence scores for each point of the completed shape point clouds.
- Obtaining a scene representation with images
- Estimating semantic information of partially-observed objects
- Determining segmented point clouds based on semantic and geometric information
- Creating metric embeddings of segmented point clouds
- Predicting completed shape point clouds
- Generating confidence scores for each point using a probabilistic machine learning model
- Controlling robot motion based on confidence scores
- Potential Applications
- Robotics - Autonomous vehicles - Augmented reality
- Problems Solved
- Improving reliability of three-dimensional shape predictions - Enhancing object recognition in partially-observed scenes
- Benefits
- Increased accuracy in shape predictions - Efficient decision-making for robots based on confidence scores
- Potential Commercial Applications
- Enhancing Object Recognition in Robotics with Confidence Scores
- Potential Commercial Applications
- Possible Prior Art
There may be prior art related to probabilistic machine learning models for object recognition and shape prediction in computer vision applications.
- Unanswered Questions
1. How does the method handle complex and cluttered scenes with multiple objects? 2. What are the limitations of the probabilistic machine learning model in predicting shape point clouds accurately?
Original Abstract Submitted
The present disclosure provides methods, apparatuses, and computer-readable mediums for evaluating a reliability of three-dimensional shape predictions. In some embodiments, a method includes obtaining a scene representation including one or more images, estimating semantic information of partially-observed objects in the scene representation, based on geometric information extracted from the one or more images, determining segmented pointclouds of the partially-observed objects based on the semantic information and the geometric information, creating metric embeddings of the segmented pointclouds corresponding to the partially-observed objects, predicting completed shape pointclouds of the partially-observed objects, generating, using a probabilistic machine learning model, confidence scores for each point of the completed shape pointclouds, based on a correlation between the geometric information and the semantic information, and controlling a motion of a robot based on the confidence scores for each point of the completed shape pointclouds.
