Image Capture Model Selection Across Detection Granularities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image detection systems lack the ability to efficiently switch between different granularities of object detection models, making it difficult for users to accurately select and perform processes on specific objects or parts within an image.
Innovation Solution
An information processing apparatus that switches between trained models with varying detection granularities, allowing users to select and perform processes like tracking, autofocus, or counting on objects or specific parts by displaying detection frames and utilizing user input to determine the process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple trained models with different detection granularities are maintained, then detection flexibility and accuracy are improved, but device complexity increases
Solution Approach 1:
The information processing apparatus is designed to perform multiple detection functions using a single system by switching between different trained models. The system can detect objects at various granularities (e.g., entire objects vs. specific parts) using the same hardware platform, making the system multi-functional without requiring separate detection systems for each granularity level.
Solution Approach 2:
The system dynamically switches between different trained models based on user selection and detection needs. Instead of using a static detection model, the system can adaptively change the detection granularity by selecting appropriate pre-trained models, allowing flexible adjustment of detection precision without physical reconfiguration.
2Ease of operation
If users can select from multiple detection granularities, then ease of operation is improved, but device complexity increases
Solution Approach 1:
The system provides self-service functionality by automatically managing multiple trained models and presenting them to users in an intuitive manner. The apparatus handles model selection, switching, and management automatically, allowing users to benefit from multiple detection granularities without needing to understand the underlying complexity of managing multiple models.
Solution Approach 2:
The system acts as an intermediary between the complex world of multiple detection models and the user. It manages the plurality of trained models with different granularities and presents simplified selection options to users, mediating the complexity so users can easily select desired detection granularity without dealing with model management details.
3Loss of information
If integrated detection results are provided, then information completeness is improved, but ease of operation deteriorates due to information overload
Solution Approach 1:
The system segments detection information by presenting results in hierarchical layers. Users can first view integrated detection results that show all detected objects, then selectively focus on specific objects or parts by switching to appropriate trained models. This segmentation allows users to access complete information when needed while maintaining ease of operation by allowing selective focus on relevant details.
Solution Approach 2:
The information presentation is dynamic, allowing users to switch between integrated detection results and specific object-focused results based on their needs. The system can adaptively change the level of information detail displayed, providing comprehensive information when users need complete oversight while simplifying the interface when users need to focus on specific targets for operations like autofocus or tracking.
Data Source
AI summary
A display unit switches which trained model, among a plurality of trained models whose granularities of detection for an object to be detected from an image are different from each other, is a trained model of interest, and displays on a screen a result of detection by the trained model of interest. A determination unit determines an object on which a predetermined process is to be performed based on a user operation on the result of detection by the trained model of interest.