Image Recognition Using Partial Image Segmentation for Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in image recognition is the increased possibility of false recognition when distinguishing between many objects, due to the small differences in their feature amounts.
Innovation Solution
An image recognition apparatus that includes a first recognition unit for identifying an object's position, a partial image extraction unit for extracting a region associated with the recognized object, and a second recognition unit using a selected learned model to recognize additional objects within the partial image, which can be resized or deformed based on specific rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If many kinds of objects are intended to be discriminatingly recognized by image recognition, then the recognition scope is improved, but the possibility of false recognition is increased because the difference among respective feature amounts becomes small
Solution Approach 1:
The image is divided into multiple partial images based on detected object positions. Each partial image is then processed by a learned model specialized for that object category, allowing for more precise recognition without compromising overall recognition scope.
Solution Approach 2:
Different learned models are assigned to different partial images based on the object categories they contain. This allows each region to be processed by a model optimized for its specific object type, improving recognition reliability while maintaining the ability to recognize multiple object kinds across the entire image.
2Device complexity
If a single learned model is used for recognizing multiple object kinds, then the device complexity is reduced, but the recognition precision decreases due to small feature amount differences
Solution Approach 1:
The image processing is segmented into multiple stages: first detecting object positions, then dividing the image into partial images, and finally applying appropriate learned models to each partial image. This segmentation enables precise recognition without requiring a single complex model to handle all object types simultaneously.
Solution Approach 2:
Object position detection is performed as a preliminary action before applying the learned models. This preliminary step allows the system to prepare and select the appropriate learned models in advance, ensuring that each object category is processed by the most suitable model for maximum recognition precision.
3Productivity
If the entire image is processed for object recognition, then the productivity is improved, but the measurement precision decreases when focusing on specific object categories
Solution Approach 1:
The image is segmented into partial images corresponding to different object positions and categories. This segmentation allows the system to process each partial image with a specialized learned model, improving recognition precision for specific object categories while maintaining overall processing efficiency through parallel processing of multiple partial images.
Solution Approach 2:
Each partial image is processed with a learned model specifically optimized for its object category. This local quality approach ensures that each region receives the most appropriate processing, improving precision for specific object categories while the overall system maintains high productivity through efficient resource allocation.
Data Source
AI summary
Provided are an image recognition apparatus, an image recognition method, and a program for enabling recognition of many kinds of objects with high precision. An overall recognition unit executes, for at least one given object, a process of recognizing the position of the object in an image. A partial image extraction unit extracts, from the image, a partial image which is a part of the image associated with the recognized position. A partial recognition unit executes a process of recognizing what is one or more objects represented by the partial image, the one or more objects including an object other than the given object the position of which is recognized.