Neural Network Image Segmentation for Automatic Individual Selection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional digital image editing systems are tedious and inaccurate for users to manually select and segregate individuals from backgrounds in digital images, leading to frustration and poor results.

Innovation Solution

The use of deep learning techniques, specifically a trained neural network with position and shape input channels, to automatically select individuals in digital images, improving accuracy and reducing user interaction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional digital image editing systems are used to manually select and segregate individuals from backgrounds, then users can perform image editing tasks, but the process becomes tedious and time-consuming with poor accuracy

Engineering Contradiction:

Improveaccuracy of individual selectionVSAvoidtime and effort required for manual selection

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces manual mechanical operations (users manually tracing boundary lines or selecting points) with an automated neural network system that processes images computationally. The neural network automatically identifies and segments individuals in digital images, eliminating the need for tedious manual user interactions while achieving superior accuracy compared to conventional systems.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If conventional digital image editing systems require repeated user selections of points or areas, then users can guide the selection process, but the complexity of operation increases and user frustration grows

Engineering Contradiction:

Improveease of individual selectionVSAvoidcomplexity of selection process

Core Design Contradiction:

Ease of operationVSDevice complexity

Solution Approach 1:

The neural network system performs the selection task autonomously without requiring user guidance through multiple point selections or boundary tracing. The system processes the image and automatically generates the segmentation, serving itself to complete the task that would otherwise require complex user interaction sequences.

Inventive Principle:

Principle #25Self-service

3Reliability

If conventional systems fail to accurately segregate individuals from background pixels, then manual effort is wasted, but the reliability of the selection result deteriorates

Engineering Contradiction:

Improvereliability of segregation resultVSAvoidtime spent on inaccurate selections

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces unreliable manual selection processes with a neural network-based automated system that consistently achieves accurate segmentation. The neural network has been trained to reliably distinguish individuals from background pixels, eliminating the reliability issues that plague conventional systems while reducing the time users would spend correcting inaccurate manual selections.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS9978003B2Utilizing deep learning for automatic digital image segmentation and stylization

Publication Date: 2018.05.22 ADOBE INC

AI summary

Systems and methods are disclosed for segregating target individuals represented in a probe digital image from background pixels in the probe digital image. In particular, in one or more embodiments, the disclosed systems and methods train a neural network based on two or more of training position channels, training shape input channels, training color channels, or training object data. Moreover, in one or more embodiments, the disclosed systems and methods utilize the trained neural network to select a target individual in a probe digital image. Specifically, in one or more embodiments, the disclosed systems and methods generate position channels, training shape input channels, and color channels corresponding the probe digital image, and utilize the generated channels in conjunction with the trained neural network to select the target individual.