DeepJEB++: Foundation Model-Driven Large-Scale 3D Engineering Dataset via 2D Latent Space Augmentation
Researchers have developed DeepJEB++, a novel framework that uses foundation models to generate a large-scale 3D engineering dataset. The system augments a small seed set of jet engine brackets into over 15,000 simulation-labeled 3D models. This is achieved by first augmenting in a 2D latent space using a diffusion model, then lifting the validated designs to 3D meshes with a generative foundation model, and finally assigning physics-based labels automatically. AI
IMPACT Enables creation of large-scale, simulation-labeled 3D datasets for engineering AI research.