Back to results
US 12,755,388 B2Granted
Image segmentation using one or more neural networks
AssigneeNVIDIA Corporation
InventorsDong Yang, Holger Roth, Xiaosong Wang, Ziyue Xu, Andriy Myronenko, Daguang Xu
Life of this patent
Estimating the term…
Abstract
Apparatuses, systems, and techniques are presented to predict segmentations for objects in images. In at least one embodiment, a neural network is trained to determine one or more segmentation masks corresponding to one or more objects of one or more digital images based, at least in part, on one or more boundary regions of the one or more objects.
Claims41
One or more processors, comprising:
- circuitry to:
- evaluate a boundary enhancement loss term and a segmentation loss term of a loss function, wherein:
- the boundary enhancement loss term comprises one or more loss values corresponding to one or more object boundary pixels and is determined based on a difference between one or more filtered predicted segmentation masks and one or more filtered ground truth segmentation masks;
- and
- the segmentation loss term comprises one or more loss values corresponding to one or more pixels within one or more objects;
- and
- optimize one or more neural networks based, at least in part, on the evaluation.
4 more elements
A system comprising:
- one or more processors to evaluate a boundary enhancement loss term and a segmentation loss term of a loss function, wherein;
- the boundary enhancement loss term comprises one or more loss values corresponding to one or more object boundary pixels and is determined based on a difference between one or more filtered predicted segmentation masks or one or more filtered ground truth segmentation masks;
- and
- the segmentation loss term comprises one or more loss values corresponding to one or more pixels within one or more objects;
- and
- optimize one or more neural networks based, at least in part, on the evaluation.
3 more elements
A method comprising:
- evaluating a boundary enhancement loss term and a segmentation loss term of a loss function, wherein:
- the boundary enhancement loss term comprises one or more loss values corresponding to one or more object boundary pixels and is determined based on a difference between one or more filtered predicted segmentation masks and one or more filtered ground truth segmentation masks;
- and
- the segmentation loss term comprises one or more loss values corresponding to one or more pixels within one or more objects;
- and
- optimizing one or more neural networks based, at least in part, on the evaluation.
3 more elements
A non-transitory machine-readable medium having stored thereon a set of instructions, which if performed by one or more processors, cause the one or more processors to at least:
- evaluate a boundary enhancement loss term and a segmentation loss term of a loss function, wherein:
- the boundary enhancement loss term comprises one or more loss values corresponding to one or more object boundary pixels and is determined based, at least in part, on a difference between one or more filtered predicted segmentation masks and one or more filtered ground truth segmentation masks;
- and
- the segmentation loss term comprises one or more loss values corresponding to one or more pixels within one or more objects;
- and
- optimize one or more neural networks based, at least in part, on the evaluation.
3 more elements
A processor, comprising:
- circuitry to:
- evaluate a boundary enhancement loss term and a segmentation loss term of a loss function, wherein the boundary enhancement loss term comprises one or more loss values corresponding to one or more object boundary pixels and is determined based on a difference between one or more filtered predicted segmentation masks and one or more filtered ground truth segmentation masks;
- and
- the segmentation loss term comprises one or more loss values corresponding to one or more pixels within one or more objects;
- and
- optimize one or more neural networks based, at least in part, on the evaluation.
3 more elements
A system comprising:
- one or more processors to:
- evaluate a boundary enhancement loss term and a segmentation loss term of a loss function, wherein:
- the boundary enhancement loss term comprises one or more loss values corresponding to one or more object boundary pixels and is determined based on a difference between one or more filtered predicted segmentation masks and one or more filtered ground truth segmentation masks;
- and
- the segmentation loss term comprises one or more loss values corresponding to one or more pixels within one or more objects;
- and
- optimize one or more neural networks based, at least in part, on the evaluation.
4 more elements
Dependencies are inferred from the claim text.