A method and system for constructing an image dataset of ancient paintings.
By employing adaptive compression and multimodal model selection methods, the problems of inconsistent formats, discrepancies between images and text, and redundancy in the construction of ancient painting image datasets were solved, achieving efficient and automated data processing and improved model training performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU GONGSHU DISTRICT HOLOGRAPHIC INTELLIGENT TECHNOLOGY RESEARCH INSTITUTE
- Filing Date
- 2025-12-12
- Publication Date
- 2026-05-26
AI Technical Summary
Existing ancient painting image datasets suffer from problems such as inconsistent data formats and sizes, discrepancies between images and text, severe data noise, and high redundancy, resulting in high storage costs, low processing efficiency, and poor model training performance.
An adaptive compression, multimodal model screening, and two-stage deduplication method are adopted, including adaptive compression processing, multi-dimensional automated screening, and two-stage duplicate image detection, to solve the problems of image format uniformity, image-text consistency, off-screen interference, and data redundancy.
It has enabled the efficient and automated construction of ancient painting image datasets, improving data processing efficiency and quality, reducing storage costs, and enhancing the accuracy and efficiency of model training.
Smart Images

Figure CN121305268B_ABST
Abstract
Citation Information
Patent Citations
Data deduplication method and system based on incremental calculation and feature clustering and electronic equipment
CN120470241A
Vision-language model image-text pair accurate evaluation data construction method based on optimization algorithm
CN120783147A
Multi-modal large model-based seal detection and identification method and system
CN120913186A