Adaptive Learning Model Selection for Image Capture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image capturing devices, such as life-log cameras, often capture unwanted videos when set to automatic mode, failing to capture desired moments due to changing user preferences and environments, as they do not adapt to mood or situational changes.
Innovation Solution
An image processing method and apparatus that selects a learning model based on user instructions, evaluation results, environmental conditions, and score analysis to determine whether captured images meet reference criteria, allowing for adaptive and user-intention-driven image recording.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automatic image capturing is performed at a predetermined time interval, then the device can continuously record scenes without user operation, but it captures unwanted videos and fails to capture desired moments due to changing user preferences
Solution Approach 1:
The system dynamically switches between multiple learning models (first learning model for general scenes, second learning model for specific scenes) based on real-time scene recognition. This allows the automatic image capturing function to adapt to changing user preferences and environmental conditions, improving capture accuracy while maintaining automation.
Solution Approach 2:
The system changes the parameters of the learning model being used based on scene characteristics. When a specific scene is detected, the system switches from the first learning model to the second learning model, effectively changing the evaluation parameters to match user preferences for that particular scene type.
2Device complexity
If a single learning model is used for automatic image capturing, then the device structure is simple, but it cannot adapt to different situations and changing user preferences
Solution Approach 1:
The system employs multiple learning models where the first learning model handles general scene evaluation and the second learning model handles specific scene evaluation. This multi-functional approach allows the device to adapt to various situations and user preferences without requiring a completely different system for each scenario.
Solution Approach 2:
The scene recognition unit acts as an intermediary that determines which learning model to use. It analyzes the current scene and selects the appropriate learning model (first or second) based on the detected scene type, enabling the system to adapt to different situations while maintaining a structured approach.
3Adaptability or versatility
If the learning model continuously learns from user feedback, then it can adapt to user preferences, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary classification of scenes using the scene recognition unit before applying the full learning model evaluation. This preliminary action allows the system to quickly identify specific scenes and switch to the appropriate learning model, reducing unnecessary processing time for general scenes while maintaining adaptability.
Solution Approach 2:
The learning process is segmented into different learning models (first learning model for general scenes, second learning model for specific scenes). This segmentation allows the system to apply computational resources selectively, using the more complex second learning model only when appropriate, thereby reducing overall processing time while maintaining adaptability to user preferences.
Data Source
AI summary
An image processing method includes the steps of performing processing of selecting a learning model from a plurality of learning models that have learned a reference used to record an image generated by an image sensor; performing, using the selected learning model, determination processing of determining whether the image generated by the image sensor satisfies the reference; and recording the image in a case in which it is determined that the image generated by the image sensor satisfies the reference, wherein the processing of selecting the learning model is performed based on at least one of an image capturing instruction by a user, an evaluation result of the image by the user, an environment when the image is generated, and a score of each of the learning models for the image generated by the image sensor.