面向多场景鸟类图像的垂直大模型文案生成方法
By preprocessing bird images and validating them with a validation set, the generation parameters are corrected, solving the problems of low reliability and efficiency in existing text generation technologies, and achieving efficient and accurate bird image text generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIHOME TECHNOLOGY CO LTD
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies do not consider periodically correcting generation parameters based on a validation set, which affects the reliability and efficiency of bird image caption generation.
By preprocessing the acquired images, extracting bird and background features, retrieving references from the vertical large model, identifying valid sentences, and periodically validating the generated copy through a validation set, the generation parameters are corrected to ensure the accuracy and structural rationality of the copy.
It improves the reliability and efficiency of generated copy, ensures that each sentence has a semantic anchor point and a reasonable structure, reduces performance drift during long-term operation, and reduces the false judgment rate.
Smart Images
Figure CN122113869B_ABST