From blueprint to quote in minutes: a lightweight CLIP-based solution for drawing similarity search

In the manufacturing industry, blueprints are critical for product development and cost estimation. However, when clients submit new blueprints for quotation, sales personnel often spend at least 2 days manually searching for similar archived designs to estimate cost, resulting in delayed communication. This study presents a lightweight AI-based solution utilizing the CLIP (Contrastive Language–Image Pretraining) model to accelerate drawing similarity search and quotation response. We construct a searchable vector database of historical engineering drawings and implement a blueprint quoting system powered by CLIP. Our system was developed with Python, demonstrating that similar drawings can be retrieved within 2 minutes with high semantic relevance. The developed system achieves low deployment cost and industry readiness, offering a practical AI-assisted service for drawing retrieval and quotation. This approach significantly reduces time and labor costs, and has high potential for being adopted in manufacturing enterprises seeking quotation automation.