Text-to-3D Generation for Digital Twins in Food Industry: A Point-E Model Application

This draft describes an approach to generating three-dimensional (3D) representations of food industry-related machinery and packaging from textual descriptions using the Point-E model. The motivation lies in integrating such assets into digital twins for food production enterprises, aiming to automate and accelerate 3D model creation, thereby reducing modeling costs and enhancing simulation fidelity. The study focuses on generating models for food industry machinery and packaging. Point-E transforms textual prompts into point clouds, then meshes. Results indicate that while prompt engineering and post-processing are crucial for detailed designs, the approach significantly benefits digital twin workflows. Key findings highlight trade-offs in model accuracy, detail, and polygonal complexity, alongside strategies for output improvement through iterative prompting and targeted mesh optimization.