Dynamic Learned Model Replacement for Image Capturing Apparatus Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image capturing apparatuses face storage capacity limitations when dealing with multiple learned models for various objects and image processing tasks, making it impossible to store all required models, especially when the number of object types and processing tasks increases.
Innovation Solution
An image capturing apparatus and an information processing apparatus communicate to dynamically replace learned models by acquiring and transmitting learned coefficient parameters based on history data, allowing the image capturing apparatus to change settings and adapt to different processing tasks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple learned models are stored in the image capturing apparatus to support various object types and processing tasks, then the adaptability and versatility of the apparatus is improved, but the storage capacity is exceeded
Solution Approach 1:
The patent extracts only the essential learned coefficient parameters from complete learned models and stores them in the image capturing apparatus. The information processing apparatus retains the full learned models and transmits only the necessary coefficient parameters to the image capturing apparatus, reducing storage requirements while maintaining functionality
Solution Approach 2:
The image capturing apparatus uses a single storage location to hold learned coefficient parameters that can serve multiple processing tasks including AF, AE, AWB, noise reduction, and image processing. This universal parameter storage approach allows one storage space to support multiple functions that would otherwise require separate models
2Adaptability or versatility
If learned models are replaced dynamically to accommodate different object types, then the adaptability is improved, but the device complexity increases
Solution Approach 1:
The information processing apparatus serves as an intermediary that manages the complexity of storing and managing multiple learned models. It generates history data, determines which learned coefficient parameters to transmit, and handles the replacement logic, thereby offloading complexity from the image capturing apparatus
Solution Approach 2:
The system uses history data that records captured image information and image capturing information to dynamically determine which learned coefficient parameters should be transmitted. This feedback mechanism allows the system to adaptively replace learned models based on actual usage patterns without requiring complex manual configuration
Data Source
AI summary
An image capturing apparatus capable of replacing a learned model for performing processing on a captured image, with another, as required. The image capturing apparatus generates a captured image by capturing an image of an object, communicates with an information processing apparatus storing learned models associated with a plurality of objects of different classes, respectively, executes the processing on the captured image using a learned model associated with a high-priority object, generates history data based on at least one of the captured image and image capturing information at the image capturing time, acquires learned coefficient parameters of the learned model associated with another object from the information processing apparatus based on the generated history data, and changes settings of the learned model based on the acquired learned coefficient parameters.