3D Object Segmentation and Editing for Real-Time Multi-View Scenes

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional 3D object segmentation techniques face challenges in precise segmentation of complex-shaped objects, requiring manual threshold setting, high computational complexity, and user expertise, limiting real-time processing and user-friendliness.

Innovation Solution

A method for 3D object segmentation that includes obtaining 2D images from multiple viewpoints, generating a 3D images with a SAM, and using a SAM to segment objects, followed by contrastive learning and global feature-guided learning to achieve precise and automated segmentation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual threshold setting is used for object segmentation, then segmentation accuracy may be improved, but user workload increases and ease of operation deteriorates

Engineering Contradiction:
Improvesegmentation accuracyVSAvoiduser workload
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs automatic segmentation by having the model independently determine segmentation thresholds and identify objects without requiring manual user input for threshold setting, thereby maintaining high segmentation accuracy while eliminating the need for manual adjustment

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system dynamically adjusts segmentation parameters automatically through the trained model, changing threshold values and other parameters based on the input image characteristics rather than requiring fixed manual settings, achieving both accuracy and ease of use

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If complex preprocessing or postprocessing steps are adopted, then segmentation precision may be improved, but computational complexity increases and real-time processing capability deteriorates

Engineering Contradiction:
Improvesegmentation precisionVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs necessary preprocessing operations during the training phase rather than at runtime, pre-computing feature extractions and transformations so that the actual segmentation process requires minimal additional computation, achieving high precision without real-time computational burden

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system replaces complex mechanical preprocessing steps with learned feature representations from the trained neural network, substituting manual or algorithmic feature engineering with automatic deep feature extraction that achieves comparable or superior precision with lower computational overhead

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Measurement precision

If a network structure with large computational volume is used, then segmentation accuracy may be improved, but real-time processing capability deteriorates

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidreal-time processing capability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system divides the complex segmentation task into hierarchical levels (e.g., coarse segmentation followed by fine segmentation), allowing the use of lighter network structures at each stage rather than requiring one large computationally intensive network, thereby achieving high accuracy through multiple simpler processing stages

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system employs dynamic computation strategies where the network adapts its processing depth and complexity based on the specific input characteristics, using more computational resources only when necessary to achieve the desired accuracy level, thus maintaining real-time processing capability across varying input types

Inventive Principle:
Principle #15Dynamics

4Measurement precision

If repeated user feedback is required for accurate segmentation, then segmentation accuracy may be improved, but system complexity increases and user-friendliness deteriorates

Engineering Contradiction:
Improvesegmentation accuracyVSAvoiduser-friendliness
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system incorporates automated feedback mechanisms during training where the model learns from segmentation errors and adjusts its parameters automatically, eliminating the need for repeated manual user feedback while maintaining and improving segmentation accuracy through self-correction

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The system performs self-validation and self-adjustment of segmentation results through the trained model's confidence scoring and automatic refinement capabilities, reducing or eliminating the need for user feedback loops and making the system more user-friendly while maintaining high accuracy

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP4675557A1Device, method of operating device and non-transitory recording medium
Publication Date: 2026.01.07 LG ELECTRONICS INC
  • EP4675557A1 patent drawingFigure 1
  • EP4675557A1 patent drawingFigure 2
  • EP4675557A1 patent drawingFigure 3

AI summary

A device according to one embodiment of the present disclosure may comprise a display; and at least one processor configured to: obtain a plurality of two-dimensional (2-Dimensional) images corresponding to a plurality of viewpoints, obtain a 3D image including one or more segmented 3D objects based on the obtained plurality of 2D images, display a UI (User Interface) element indicating that a segmented 3D object is selected on the display in response to a first command to select one of the one or more segmented 3D objects representing a unit object or a detailed object constituting the unit object, and display an edited 3D object on the display in response to a second command to edit the selected 3D object.