Header Mask Comparison for Raster Segmentation Retraining Timing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing workflows utilizing deep learning models face challenges in identifying when to fine-tune or retrain models due to data distribution shifts and privacy issues, particularly in workflows with multiple machine learning models, making it difficult to determine which models need retraining.

Innovation Solution

A raster retraining decision system that includes a raster segmentation model, a header mask segmentation model, and a segment comparison model to generate comparison scores, which are used to determine when to retrain the raster segmentation model, thereby improving model accuracy and reducing computational resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If machine learning models are continuously trained to handle data distribution shifts, then model performance and adaptability improve, but computational resources and time consumption increase

Engineering Contradiction:
Improvemodel adaptability to data distribution shiftsVSAvoidcomputational resources
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary actions by continuously monitoring model performance metrics and data distribution characteristics in advance, before actual retraining is needed. This allows the system to identify when retraining is necessary and when the current model suffices, avoiding unnecessary computational resources while maintaining adaptability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system implements feedback mechanisms that continuously evaluate model performance against data distribution shifts. By using performance metrics and distribution statistics as feedback, the system can dynamically decide whether retraining is needed, optimizing the balance between model adaptability and computational resource consumption.

Inventive Principle:
Principle #23Feedback

2Use of energy by moving object

If manual monitoring of model performance is used to determine retraining needs, then computational resources are saved, but time consumption and operational complexity increase

Engineering Contradiction:
Improvecomputational resourcesVSAvoidtime for model performance evaluation
Core Design Contradiction:
Use of energy by moving objectVSLoss of time

Solution Approach 1:

The system performs self-service by automatically monitoring its own performance metrics and data distribution characteristics without requiring manual intervention. The automated evaluation system continuously assesses model performance and triggers retraining decisions autonomously, eliminating time-consuming manual monitoring while optimizing computational resource usage.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system replaces manual mechanical monitoring processes with automated electronic evaluation mechanisms. By substituting human-based performance assessment with automated algorithms that continuously evaluate model metrics and data distributions, the system eliminates time consumption associated with manual checks while maintaining efficient computational resource management.

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

3Reliability

If all machine learning models in a workflow are retrained simultaneously, then overall system accuracy improves, but device complexity and operational difficulty increase

Engineering Contradiction:
Improveworkflow accuracyVSAvoidretraining management complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system applies segmentation by evaluating and retraining individual machine learning models independently based on their specific performance metrics and data distribution characteristics. Rather than treating all models as a monolithic unit, the system segments the retraining process to focus only on models that actually need updates, reducing operational complexity while maintaining overall workflow accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system implements local quality by applying different retraining strategies to different models based on their individual performance characteristics and data distribution shifts. Each model is evaluated and managed according to its specific requirements, allowing selective retraining that improves accuracy without uniformly increasing complexity across the entire workflow.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4704053A1Raster retraining decision system
Publication Date: 2026.03.04 SERVICES PETROLIERS SCHLUMBERGER SA
  • EP4704053A1 patent drawingFigure 1
  • EP4704053A1 patent drawingFigure 2
  • EP4704053A1 patent drawingFigure 3

AI summary

A method implements a raster retraining decision system. The method includes executing a raster segmentation model for a first stage to generate multiple masks including a first header mask. The method further includes executing a header mask segmentation model for a second stage to generate a second header mask. The method further includes executing a segment comparison model using the first header mask with the second header mask to generate a comparison score. The method further includes generating a raster retraining score from the comparison score for the raster segmentation model of a raster digitization engine. The method further includes retraining the raster segmentation model using the raster retraining score.