Enhanced flowchart transformation and code generation using LLM models
Akshat Johri, Saurabh Sharma, Vinay Chaudhary, Gaurav Raj, Shiraz Khurana · 2025
This paper explores the automated conversion of flowchart images into UML and class diagrams by integrating advanced AI techniques. While visual modeling tools like flowcharts, UML diagrams, and class diagrams are crucial in software engineering. To address this, an AI-based tool is introduced to translate the conceptual workflows to structured software models. The proposed system, leverages transformer models, to identify key elements within flowchart images, such as shapes, connectors, and labeling. Precisely, the system utilizes GPT-3.5, which is a powerful transformer model, to analyze the visual information and deduce relationships between the components of the flowchart. Based on the extracted points, the tool reconstructs clean, digital UML or class diagrams aligned with standard modeling contracts. The analysis and potential enhance the system&s;s capabilities to consider other models like Llama and Claude. The system architecture includes modules for image preprocessing, component recognition, relationship mapping, and diagram rendering. Users have to upload flowchart images, and the tool processes the input to generate an editable and exportable version of the diagram. Experimental results demonstrate the achievement of high accuracy in diagram conversion, offering significant time savings in model creation. This tool is very beneficial for software developers, educators, and students seeking to design documentation and maintain consistency across modeling phases. Future enhancements will include real-time sketch recognition, multi-diagram support, and integration with popular design tools.