Photo management
First Claim
Patent Images
1. A method for image processing, comprising:
- determining, with a first pre-trained classifier, features of a plurality of stored images, the plurality of stored images being different from a plurality of training images used for training the first pre-trained classifier;
assigning each image of the plurality of stored images to a cluster of a plurality of clusters based at least in part on the determined features;
assigning a label to an image of the plurality of stored images based on a distance of the image to a closest cluster center corresponding to the image being less than a first threshold, a pre-defined number of images associated with the closest cluster center being less than a second threshold, and/or a percentage of images associated with the closest cluster center being less than a third threshold;
modifying the assigned label;
re-training the first pre-trained classifier with the image when the assigned label is modified; and
re-sorting the image based at least in part on the re-trained classifier.
1 Assignment
0 Petitions
Accused Products
Abstract
A method for image processing includes determining features of multiple stored images from a pre-trained deep convolutional network. The method also includes clustering each image of the multiple stored images based on the determined features.
72 Citations
33 Claims
-
1. A method for image processing, comprising:
-
determining, with a first pre-trained classifier, features of a plurality of stored images, the plurality of stored images being different from a plurality of training images used for training the first pre-trained classifier; assigning each image of the plurality of stored images to a cluster of a plurality of clusters based at least in part on the determined features; assigning a label to an image of the plurality of stored images based on a distance of the image to a closest cluster center corresponding to the image being less than a first threshold, a pre-defined number of images associated with the closest cluster center being less than a second threshold, and/or a percentage of images associated with the closest cluster center being less than a third threshold; modifying the assigned label; re-training the first pre-trained classifier with the image when the assigned label is modified; and re-sorting the image based at least in part on the re-trained classifier. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16)
-
-
17. An apparatus for image processing, the apparatus comprising:
-
a memory unit; and at least one processor coupled to the memory unit, the at least one processor being configured; to determine, with a first pre-trained classifier, features of a plurality of stored images, the plurality of stored images being different from a plurality of training images used for training the first pre-trained classifier; to assign each image of the plurality of stored images to a cluster of a plurality of clusters based at least in part on the determined features; to assign a label to an image of the plurality of stored images based on a distance of the image to a closest cluster center corresponding to the image being less than a first threshold, a pre-defined number of images associated with the closest cluster center being less than a second threshold, and/or a percentage of images associated with the closest cluster center being less than a third threshold; to modify the assigned label; to re-train the first pre-trained classifier with the image when the assigned label is modified; and to re-sort the image based at least in part on the re-trained classifier. - View Dependent Claims (18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31)
-
-
32. A non-transitory computer-readable medium having program code recorded thereon for image processing, the program code being executed by a processor and comprising:
-
program code to determine, with a first pre-trained classifier, features of a plurality of stored images, the plurality of stored images being different from a plurality of training images used for training the first pre-trained classifier; program code to assign each image of the plurality of stored images to a cluster of a plurality of clusters based at least in part on the determined features; program code to assign a label to an image of the plurality of stored images based on a distance of the image to a closest cluster center corresponding to the image being less than a first threshold, a pre-defined number of images associated with the closest cluster center being less than a second threshold, and/or a percentage of images associated with the closest cluster center being less than a third threshold; program code to modify the assigned label; program code to re-train the first pre-trained classifier with the image when the assigned label is modified; and program code to re-sort the image based at least in part on the re-trained classifier.
-
-
33. An apparatus for image processing, the apparatus comprising:
-
means for determining, with a first pre-trained classifier, features of a plurality of stored images, the plurality of stored images being different from a plurality of training images used for training the first pre-trained classifier; means for assigning each image of the plurality of stored images to a cluster of a plurality of clusters based at least in part on the determined features; means for assigning a label to an image of the plurality of stored images based on a distance of the image to a closest cluster center corresponding to the image being less than a first threshold, a pre-defined number of images associated with the closest cluster center being less than a second threshold, and/or a percentage of images associated with the closest cluster center being less than a third threshold; means for modifying the assigned label; means for re-training the first pre-trained classifier with the image when the assigned label is modified; and means for re-sorting the image based at least in part on the re-trained classifier.
-
Specification