Elastic distortions for automatic generation of labeled data
First Claim
Patent Images
1. A system that facilitates training a classifier, comprising:
- a component that receives a data set to be employed in connection with training the classifier; and
an expansion component that applies elastic distortion algorithm(s) to a subset of the data set to generate additional training data.
2 Assignments
0 Petitions
Accused Products
Abstract
A system that facilitates generation of data that can be employed in connection with training a classifier. The system comprises a component that receives a data set that is employed in connection with training the classifier, and an expansion component that applies elastic distortion algorithm(s) to a subset of the data set to generate additional labeled training data.
-
Citations
75 Claims
-
1. A system that facilitates training a classifier, comprising:
-
a component that receives a data set to be employed in connection with training the classifier; and
an expansion component that applies elastic distortion algorithm(s) to a subset of the data set to generate additional training data. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 68, 72)
-
-
39. A method for generating data that can be utilized in connection with training a classifier employed to analyze at least one of handwriting, speech, and or objects, comprising:
-
receiving a data set formatted as one or more image(s) with pixel(s) that define the one or more image(s);
generating a smooth displacement field using randomly generated numbers; and
elastically distorting the image(s) based at least in part upon the displacement field. - View Dependent Claims (40, 41, 42, 43, 44, 45, 46, 50, 51, 52, 53, 66, 69)
-
-
47. A method for generating data that can be utilized in connection with training a classifier employed to analyze at least one of handwriting, speech, and or objects, comprising:
-
receiving a data set formatted as one or more trajectory(ies) that define one or more pattern(s) to classify;
associating time value(s) with coordinates of point(s) associated with the trajectory(ies), thereby enabling description of the one or more trajectory(ies) by functions x(t) and y(t);
generating a smooth displacement field via using randomly generated numbers;
adding the smooth displacement field to the functions x(t) and y(t) to generate new elastically distorted trajectories x(t) and y(t). - View Dependent Claims (48, 49, 70, 73)
-
-
54. A system that facilitates generation of labeled data employed in connection with training a classifier, comprising:
-
means for receiving a data set formatted as one or more image(s), the one or more image(s) defined by one or more pixel(s); and
means for elastically distorting the one or more image(s). - View Dependent Claims (55, 56, 71, 74)
-
-
57. A system that facilitates generation of labeled data employed in connection with training a classifier, comprising:
-
means for receiving a data set formatted as one or more trajectories, the one or more trajectory(ies) defined by one or more point(s); and
means for elastically distorting the one or more trajectory(ies). - View Dependent Claims (58, 59, 60, 61, 62, 63, 64, 65)
-
-
67. A data packet that passes between at least two computer processes, comprising:
a field that stores an elastically distorted image, the image distorted via application of a smooth displacement field to an initial image, the displacement field generated at least in part by random values output from a random number generator.
-
75. A portable computing device, comprising:
-
a component that receives a human generated input; and
a classifier that was trained via employment of an expansion component that applies elastic distortion algorithm(s) to a subset of a training data set to generate additional training data.
-
Specification