Back to results
US 11,238,310 B2Granted
Training data acquisition method and device, server and storage medium
AssigneeBaidu Online Network Technology (Beijing) Co., Ltd.
InventorsSu Li, Libo Fu
Life of this patent
Estimating the term…
Abstract
A training data acquisition method and device, a server and a storage medium are provided. The training data acquisition method is applied to a classifier and includes the following steps: obtaining an image search target according to an input of a user; providing images to the user according to the image search target, to display the images; and selecting at least one image from the displayed images, and determining a target-classification pair as training data according to the at least one image; where the target-classification pair includes the image search target and an entity-based classification of the at least one image.
Claims15
A training data acquisition method, comprising:
- receiving an image search target from a computer;
- sending images to the computer according to the image search target, to display the images on the computer;
- and
- selecting at least one image from the displayed images and determining a target-classification pair as training data according to the at least one image;
- wherein the target-classification pair comprises the image search target and an entity-based classification of the at least one image,
- wherein the selecting at least one image from the displayed images and determining a target-classification pair as training data comprises:
- determining an image related to an untrusted click;
- filtering out the image related to the untrusted click from the at least one image, and determining the target-classification pair according to a filtering result;
- and
- evaluating accuracy of the target-classification pair, and determining the training data according to the accuracy, and
- wherein the determining an image related to an untrusted click comprises:
- clustering the at least one image into at least one classification, and determining a click on an image as the untrusted click, in a case that the number of images in the classification of the clicked image does not satisfy a predetermined condition, or in a case that a difference and/or ratio between the number of images in the classification of the clicked image and that in a classification with the largest number of images does not satisfy a predetermined condition;
- and
- classifying the at least one image into at least one classification based on a visual entity, and determining a click on an image as the untrusted click, in a case that the number of images in the visual-entity-based classification of the clicked image does not satisfy a predetermined condition, or in a case that a difference and/or ratio between the number of images in the visual-entity-based classification of the clicked image and that in a visual-entity-based classification with the largest number of images does not satisfy a predetermined condition.
11 more elements
A training data acquisition device, comprising:
- one or more processors;
- and
- a storage device configured for storing one or more programs, wherein
- the one or more programs are executed by the one or more processors to enable the one or more processors to perform:
- receiving an image search target from a computer;
- sending images to the computer according to the image search target, to display the images on the computer;
- and
- selecting at least one image from the displayed images and determining a target-classification pair as training data according to the at least one image;
- wherein the target-classification pair comprises the image search target and an entity-based classification of the at least one image,
- wherein the one or more programs, when executed by the one or more processors, enable the one or more processors further to perform:
- determining an image related to an untrusted click;
- filtering out the image related to the untrusted click from the at least one image, and determining the target-classification pair according to a filtering result;
- and
- evaluating accuracy of the target-classification pair, and determining the training data according to the accuracy, and
- wherein the one or more programs, when executed by the one or more processors, enable the one or more processors further to perform:
- clustering the at least one image into at least one classification, and determining a click on an image as the untrusted click, in a case that the number of images in the classification of the clicked image does not satisfy a predetermined condition, or in a case that a difference and/or ratio between the number of images in the classification of the clicked image and that in a classification with the largest number of images does not satisfy a predetermined condition;
- and
- classifying the at least one image into at least one classification based on a visual entity, and determining a click on an image as the untrusted click, in a case that the number of images in the visual-entity-based classification of the clicked image does not satisfy a predetermined condition, or in a case that a difference and/or ratio between the number of images in the visual-entity-based classification of the clicked image and that in a visual-entity-based classification with the largest number of images does not satisfy a predetermined condition.
15 more elements
A non-transitory computer readable storage medium storing a computer program, wherein the program is executed by a processor to perform:
- receiving an image search target from a computer;
- sending images to the computer according to the image search target, to display the images on the computer;
- and
- selecting at least one image from the displayed images and determining a target-classification pair as training data according to the at least one image;
- wherein the target-classification pair comprises the image search target and an entity-based classification of the at least one image,
- wherein the selecting at least one image from the displayed images and determining a target-classification pair as training data comprises:
- determining an image related to an untrusted click;
- filtering out the image related to the untrusted click from the at least one image, and determining the target-classification pair according to a filtering result;
- and
- evaluating accuracy of the target-classification pair, and determining the training data according to the accuracy, and
- wherein the determining an image related to an untrusted click comprises:
- clustering the at least one image into at least one classification, and determining a click on an image as the untrusted click, in a case that the number of images in the classification of the clicked image does not satisfy a predetermined condition, or in a case that a difference and/or ratio between the number of images in the classification of the clicked image and that in a classification with the largest number of images does not satisfy a predetermined condition;
- and
- classifying the at least one image into at least one classification based on a visual entity, and determining a click on an image as the untrusted click, in a case that the number of images in the visual-entity-based classification of the clicked image does not satisfy a predetermined condition, or in a case that a difference and/or ratio between the number of images in the visual-entity-based classification of the clicked image and that in a visual-entity-based classification with the largest number of images does not satisfy a predetermined condition.
11 more elements
Dependencies are inferred from the claim text.