Dual Recognition Model Integration for Stable Subject Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing object detection systems face challenges in maintaining detection performance when additional learning destabilizes successful detection of subjects, particularly in face detection, leading to unpredictable recognition accuracy.

Innovation Solution

An information processing apparatus integrates a fixed model and a customizable custom model to maintain detection performance by combining their results using a weighted integration method, ensuring the fixed model's accuracy is preserved while allowing user-specific customization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If additional learning is performed to improve detection accuracy of a specific subject, then detection accuracy for that specific subject is improved, but detection performance for other subjects deteriorates

Engineering Contradiction:

Improvedetection accuracyVSAvoiddetection stability

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The patent divides the single recognition model into two separate models: a fixed model that maintains original detection performance and a custom model that performs additional learning for specific subjects. This segmentation allows each model to specialize in different aspects without interfering with each other's performance.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent combines the fixed model and custom model into an integrated system that processes detection requests. The integration unit coordinates both models and merges their detection results, allowing the system to benefit from both the stability of the fixed model and the specialized accuracy of the custom model.

Inventive Principle:

Principle #5Merging (Combining)

2Adaptability or versatility

If a customizable recognition model is trained for user-specific detection needs, then user-specific detection accuracy is improved, but the original detection performance is compromised

Engineering Contradiction:

Improvecustomization capabilityVSAvoidoriginal detection accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent separates the recognition functionality into a fixed model that preserves original detection accuracy and a custom model that provides user-specific customization. This segmentation enables both adaptability and preservation of original performance simultaneously.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The integration unit acts as an intermediary that manages the interaction between the fixed model and custom model. It determines when to use each model and how to combine their results, ensuring that customization needs are met while maintaining original detection capabilities.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20260024322A1Information processing apparatus, control method of information processing apparatus, and storage medium

Publication Date: 2026.01.22 CANON KK

AI summary

An information processing apparatus that detects a subject from an input image, the information processing apparatus comprising: a storage unit that stores a fixed model that is a non-changeable recognition model learned so as to detect a subject of a predetermined category and a custom model that is a customizable recognition model learned so as to detect a subject in an identical category to the fixed model; a setting unit that sets an integration method of a detection result using the fixed model and a detection result using the custom model; and an integration unit that acquires an integration detection result by integrating, based on the integration method, each detection result to the input image.