Guiding Network Model for Image Description Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems generate low-quality descriptive statements due to inaccurate guiding information, which affects the accuracy and quality of the generated descriptions.
Innovation Solution
A guiding network model is introduced between the encoder and the decoder to generate guiding information based on the annotation vector set, improving the accuracy and quality of the descriptive statements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If artificially designed guiding information is extracted from the image, then the encoding process can be guided to generate descriptive statements, but the quality and accuracy of the generated descriptive statements are low
Solution Approach 1:
A guiding network model is introduced as an intermediary component between the encoder and decoder. This guiding network receives the encoded image features and generates guiding information that directs the decoding process. The guiding network acts as a mediator that transforms raw image features into structured guidance signals, improving the accuracy of generated descriptive statements while managing system complexity through modular architecture design.
2Measurement precision
If a guiding network model is added between encoder and decoder, then the accuracy of guiding information is improved, but the system complexity increases
Solution Approach 1:
The system is segmented into distinct functional modules: an encoder for feature extraction, a guiding network for generating guiding information, and a decoder for generating descriptive statements. This segmentation allows each component to specialize in its specific function, improving the precision of guiding information generation while managing overall system complexity through clear modular boundaries and independent optimization of each segment.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A method for recognizing an image, a terminal, and a storage medium belong to the field of machine learning. The method includes: performing feature extraction on a to-be-recognized image by using an encoder, to obtain a feature vector and a first annotation vector set (101); performing initialization processing on the feature vector to obtain first initial input data; generating first guiding information based on the first annotation vector set by using a first guiding network model, where the first guiding network model is configured to generate guiding information according to an annotation vector set of an image; and determining a descriptive statement of the image based on the first guiding information, the first annotation vector set and the first initial input data by using a decoder. A guiding network model that can generate guiding information according to an annotation vector set of an image is added between the encoder and the decoder. Therefore, the guiding information generated using the guiding network model is more accurate, thereby accurately guiding an encoding process, and improving quality of the generated descriptive statement.