US 10,853,704 B2 · Model-based image labeling and/or segmentation

Abstract

In some embodiments, reduction of computational resource usage related to image labeling and/or segmentation may be facilitated. In some embodiments, a collection of images may be used to train one or more prediction models. Based on a presentation of an image on a user interface, an indication of a target quantity of superpixels for the image may be obtained. The image may be provided to a first prediction model to cause the prediction model to predict a quantity of superpixels for the image. The target quantity of superpixels may be provided to the first model to update the first model's configurations based on (i) the predicted quantity and (ii) the target quantity. A set of superpixels may be generated for the image based on the target quantity, and segmentation information related to the superpixels set may be provided to a second prediction model to update the second model's configurations.

Claims19

Dependent claims

A method of facilitating reduction of computational resource usage related to image labeling and/or segmentation, the method being implemented by a computer system that comprises one or more processors executing computer program instructions that, when executed, perform the method, the method comprising:

obtaining a collection of images comprising 1000 or more images;
and
with respect to each image of the 1000 or more images:
causing presentation of the image on a user interface;
obtaining, based on the presentation of the image, an indication of a target quantity of superpixels for the image;
providing the image to a first neural network to cause the first neural network to predict a quantity of superpixels for the image;
and
providing the target quantity of superpixels as reference feedback to the first neural network to cause the first neural network to assess the predicted quantity of superpixels against the target quantity of superpixels, the first neural network updating one or more configurations of the first neural network based on the first neural network's assessment of the predicted quantity of superpixels.

5 more elements

A system comprising:

a computer system that comprises one or more processors programmed with computer program instructions that, when executed, cause the computer system to:
obtain an indication of a target quantity of segments for an image;
provide the image to a prediction model to cause the prediction model to predict a quantity of segments for the image;
and
provide the target quantity of segments as reference feedback to the prediction model to cause the prediction model to assess the predicted quantity of segments against the target quantity of segments, the prediction model updating one or more portions of the prediction model based on the prediction model's assessment of the predicted quantity of segments.

2 more elements

A method implemented by one or more processors executing computer program instructions that, when executed, perform the method, the method comprising:

obtaining an image related to a concept;
obtaining an indication of a quantity of superpixels for the image;
generating a set of superpixels for the image based on the quantity of superpixels for the image;
providing the image to a neural network to cause the neural network to predict one or more segments related to the concept in the image;
and
providing segmentation information related to the set of superpixels for the image as reference feedback to the neural network to cause the neural network to assess the one or more predicted segments against the segmentation information, the neural network updating one or more portions of the neural network based on the neural network's assessment of the one or more predicted segments.

3 more elements

Dependencies are inferred from the claim text.

Model-based image labeling and/or segmentation

Life of this patent

Model-based image labeling and/or segmentation

Life of this patent

Abstract

Claims19