一种平面设计图的生成方法和装置
By optimizing the generation model through reinforcement learning and a multi-dimensional reward mechanism, the problem of lack of quality assessment in existing 2D design drawing generation systems is solved, and higher quality and more stable design drawing generation is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHANGSHA GEFANG BIOMEDICAL TECHNOLOGY RESEARCH INSTITUTE CO LTD
- Filing Date
- 2026-04-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing 2D design drawing generation systems lack dynamic quality assessment and quantitative feedback mechanisms during the training process, resulting in insufficient matching between the generated design drawings and user intentions. In particular, when coordinating multiple layers of elements, problems such as confusion, overlap, or disproportion can easily occur.
The generative model is trained using reinforcement learning. The quality of the generated planar design drawings is evaluated by the evaluation model, and the reward value is calculated and fed back to the generative model to optimize its generation strategy. Combined with phased training and multi-dimensional reward mechanism, it is ensured that the generative model can generate more reasonable and stable design drawings.
The quality and stability of the generated 2D design drawings have been improved, enabling them to better conform to user text instructions. The generated design drawings have a reasonable layout, good visual effects, and harmonious text typesetting.
Smart Images
Figure CN122049121B_ABST