Learning Image Selection Using Metadata and Camera Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data management systems require manual collection of large numbers of images for AI learning, which can be effortful and may result in images that are not suitable for the intended use case.
Innovation Solution
An information processing device and method that selects learning images from a pre-held image group based on the specific use case, using a graphical user interface (GUI) to input settings and generate a data set, incorporating metadata and camera simulation to ensure image suitability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If manual collection methods are used (imaging actual scenes, searching Internet, using published data sets), then a large number of images can be collected, but it requires significant effort and the collected images may not be appropriate for the AI use case
Solution Approach 1:
The patent uses pre-held image groups as templates or copies that can be selected and reused for different AI learning purposes. Instead of manually collecting images each time, the system maintains a library of pre-collected image groups that can be copied and applied to various use cases, significantly reducing collection effort while maintaining image quantity
Solution Approach 2:
The patent performs image collection and organization in advance, creating pre-held image groups before they are needed for AI learning. This preliminary action stores diverse image collections in advance, so when a specific use case arises, the system can quickly select from pre-prepared groups rather than collecting images at the moment of need
2Quantity of substance
If manual collection methods are used, then images can be obtained, but the collected images may not be appropriate for the intended use case
Solution Approach 1:
The patent segments the overall image collection task into distinct pre-held image groups, each organized by specific characteristics or categories. This segmentation allows the system to select only the relevant group matching the intended use case, ensuring appropriateness while maintaining sufficient quantity for AI learning
Solution Approach 2:
The patent introduces pre-held image groups as an intermediary layer between general image sources and specific AI use cases. These intermediary groups are pre-organized and characterized, serving as a bridge that ensures matchin g between image characteristics and use case requirements without direct manual selection
3Adaptability or versatility
If custom image collection is performed, then images can be tailored to specific needs, but the process becomes time-consuming and complex
Solution Approach 1:
The patent performs image collection, categorization, and organization in advance into pre-held image groups with defined characteristics. This preliminary customization work is done once, allowing rapid selection and reuse for multiple different use cases without repeating the collection process, thus maintaining adaptability while reducing time loss
Solution Approach 2:
The patent creates pre-held image groups that serve multiple purposes and can be applied to various AI learning use cases. Each image group is organized with universal characteristics that make them adaptable to different applications, eliminating the need for separate custom collection processes for each use case
Data Source
AI summary
The present technique relates to an information processing device, an information processing method, and a recording medium that easily acquire an image suitable for a use case of AI. The information processing device according to the present technique includes a selection unit that selects a learning image used for learning of a learning model, according to a use case of the learning model using an image as an input, from among an image group held in advance. The present technique is applicable, for example, to a data set generation device that generates a data set including a large number of learning images.