Trademark Recognition Using Image-Text Feature Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object detection-based trademark recognition algorithms rely solely on image features, leading to poor precision in trademark recognition.
Innovation Solution
A data processing method that extracts both image and text features from an image, performs fusion on these features, and uses contrastive learning to enhance discriminability, thereby improving recognition precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing image classification networks solely rely on image features for trademark recognition, then the recognition process is simple, but the trademark recognition precision is poor
Solution Approach 1:
The patent merges image feature extraction and text feature extraction into a unified neural network architecture. The network simultaneously processes both image and text inputs through separate but integrated pathways, combining their features at multiple levels to achieve improved trademark recognition precision while maintaining architectural efficiency
Solution Approach 2:
The neural network is segmented into distinct functional modules: an image processing branch that extracts visual features, a text processing branch that extracts textual features, and a fusion module that integrates both feature types. This segmentation allows each module to specialize while contributing to the overall recognition system
2Measurement precision
If both image and text features are extracted and fused, then recognition precision is improved, but processing complexity increases
Solution Approach 1:
The patent introduces a multi-dimensional feature space by extracting features from both image and text domains. Instead of relying on a single feature dimension, the system creates a composite feature representation that incorporates visual and textual dimensions, thereby improving recognition accuracy through enhanced feature discrimination
Solution Approach 2:
The patent introduces a feature fusion module as an intermediary between the image processing and text processing branches. This intermediary component integrates features from both modalities through learned transformations, enabling the system to combine complementary information while managing processing complexity through structured fusion operations
Data Source
AI summary
A data processing method is applied to image processing. The method includes: obtaining a first image and a second image, where the first image and the second image include text; obtaining an image feature of the first image and an image feature of the second image through a first neural network; obtaining, through a second neural network, a text feature of text included in the first image and a text feature of text included in the second image; performing fusion on a first feature representation and a third feature representation to obtain a first target feature representation; performing fusion on a second feature representation and a fourth feature representation to obtain a second target feature representation; determining a loss based on a relationship between the first target feature representation and the second target feature representation; and updating the first neural network based on the loss.