Automated collection and labeling of object data

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.

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.