Hierarchical Content Recognition for Accurate Fine-Grained Labels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content recognition models face low classification accuracy when pursuing fine-grained classification results, leading to degraded generalization capability.
Innovation Solution
A method involving obtaining a first object category at a first category hierarchy, using a target category recognition model to obtain a second object category at a refined hierarchy, and adjusting the recognition result based on prediction confidence and label tree data to balance accuracy and granularity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the classification model is trained by using training samples of more specific categories to recognize fine-grained object categories, then the classification granularity is improved, but the classification accuracy of the recognition result deteriorates
Solution Approach 1:
The category hierarchy is segmented into multiple levels (first category hierarchy and second category hierarchy). The first recognition model operates at the first category hierarchy to obtain coarse-grained categories, while the second recognition model operates at the second category hierarchy to obtain fine-grained categories. This segmentation allows the system to leverage both coarse-grained generalization capability and fine-grained classification precision without the accuracy degradation that occurs when a single model pursues fine-grained classification alone.
Solution Approach 2:
The patent implements a nested structure where the second category hierarchy is embedded within the first category hierarchy. The category hierarchies are organized such that fine-grained categories (second hierarchy) are nested under coarse-grained categories (first hierarchy). This nested arrangement allows the recognition system to progressively refine category assignments from coarse to fine granularity, maintaining accuracy while achieving detailed classification.
2Measurement precision
If the classification model pursues fine-grained classification results, then the object category precision is improved, but the generalization capability deteriorates
Solution Approach 1:
The recognition system is segmented into multiple models operating at different category hierarchies. The first recognition model handles coarse-grained categories to maintain generalization capability, while the second recognition model handles fine-grained categories to achieve precision. This segmentation allows each model to specialize in its appropriate granularity level, preventing the generalization degradation that occurs when a single model is optimized for fine-grained classification.
Solution Approach 2:
The system dynamically selects between different recognition models and category hierarchies based on the specific recognition task and input characteristics. The method allows flexible switching between first and second category hierarchies, and between different recognition models, to adapt to varying requirements for precision versus generalization in different recognition scenarios.
Data Source
AI summary
Embodiments of the present disclosure provide a content recognition method and apparatus, an electronic device, and a storage medium. The content recognition method includes: obtaining a first object category of a main object in an image to be recognized based on the main object; obtaining a corresponding target category recognition model based on the first object category, and processing the image to be recognized based on the target category recognition model to obtain a second object category and a corresponding prediction confidence; and obtaining a first recognition result of the image to be recognized based on the second object category and the corresponding prediction confidence, where the first recognition result represents a predicted object category of the main object, and the predicted object category is between the first category hierarchy and the second category hierarchy.