Automated collection and labeling of object data
First Claim
1. A system for automatically collecting and labeling data about a physical object, comprising:
- a mechanical device configured to maneuver the object within a three-dimensional (3D) workspace in a real-world environment;
one or more sensors configured to capture and output images of the object when it is within one or more prescribed regions of the workspace;
a computing device; and
a computer program having program modules executable by the computing device, the computing device being directed by the program modules of the computer program to,calibrate the sensors to the coordinate system of the mechanical device,for each of a plurality of different object poses within said regions of the workspace,use the mechanical device to maneuver the object into the object pose,input an image of the object from each of the sensors while the object is in the object pose,input data specifying the object pose, the data specifying the pose being in the coordinate system of the mechanical device, andlabel the image of the object input from each of the sensors with the data specifying the object pose and with information identifying the object, andgenerate a database for the object, said database comprising the labeled image of the object input from each of the sensors for each of said different object poses.
1 Assignment
0 Petitions
Accused Products
Abstract
Data about a physical object in a real-world environment is automatically collected and labeled. A mechanical device is used to maneuver the object into different poses within a three-dimensional workspace in the real-world environment. While the object is in each different pose an image of the object is input from one or more sensors and data specifying the pose is input from the mechanical device. The image of the object input from each of the sensors for each different pose is labeled with the data specifying the pose and with information identifying the object. A database for the object that includes these labeled images can be generated. The labeled images can also be used to train a detector and classifier to detect and recognize the object when it is in an environment that is similar to the real-world environment.
-
Citations
20 Claims
-
1. A system for automatically collecting and labeling data about a physical object, comprising:
-
a mechanical device configured to maneuver the object within a three-dimensional (3D) workspace in a real-world environment; one or more sensors configured to capture and output images of the object when it is within one or more prescribed regions of the workspace; a computing device; and a computer program having program modules executable by the computing device, the computing device being directed by the program modules of the computer program to, calibrate the sensors to the coordinate system of the mechanical device, for each of a plurality of different object poses within said regions of the workspace, use the mechanical device to maneuver the object into the object pose, input an image of the object from each of the sensors while the object is in the object pose, input data specifying the object pose, the data specifying the pose being in the coordinate system of the mechanical device, and label the image of the object input from each of the sensors with the data specifying the object pose and with information identifying the object, and generate a database for the object, said database comprising the labeled image of the object input from each of the sensors for each of said different object poses. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. A system for training a detector and classifier to detect and recognize a physical object, comprising:
-
one or more computing devices, said computing devices being in communication with each other via a computer network whenever there is a plurality of computing devices; and a computer program having program modules executable by the one or more computing devices, the one or more computing devices being directed by the program modules of the computer program to, for each of a plurality of different object poses within one or more prescribed regions of a three-dimensional workspace in a real-world environment, receive an image of the object from each of one or more sensors while the object is in the object pose, said sensors being configured to capture and output images of the object when it is within said regions of the workspace, receive data specifying the object pose from a mechanical device configured to maneuver the object into the pose, the data specifying the pose being in the coordinate system of the mechanical device, the sensors being calibrated to the coordinate system of the mechanical device, and label the image of the object received from each of the sensors with the data specifying the object pose, and use the labeled image of the object received from each of the sensors for each of said different object poses to train the detector and classifier, said trained detector and classifier then being able to detect and recognize the object when it is in an environment similar to the real-world environment. - View Dependent Claims (18, 19)
-
-
20. A system for detecting and recognizing a physical object in a real-world environment, comprising:
-
a sensor configured to capture and output images of the object when it is in the real-world environment; one or more computing devices, said computing devices being in communication with each other via a computer network whenever there is a plurality of computing devices; and a computer program having program modules executable by the one or more computing devices, the one or more computing devices being directed by the program modules of the computer program to, receive an image of the object from the sensor while the object is in a particular object pose in the real-world environment, receive data specifying said particular object pose, and use a pre-trained object detector and classifier to detect and recognize the object from the image thereof and the data specifying said particular object pose, said detector and classifier having been trained using a database comprising images of the object that were captured while the object was maneuvered by a mechanical device into a plurality of different object poses in an environment similar to the real-world environment, each of the images of the object in said database being labeled with data specifying the object pose the object was in when the image thereof was captured.
-
Specification