On the Noise Scheduling for Generating Plausible Designs with Diffusion Models
This addresses the need for more reliable generative models in industries like fashion and automotive, where structural integrity is critical, though it is incremental as it builds on existing diffusion models.
The paper tackled the problem of generating structurally plausible designs with diffusion models by analyzing noise scheduling's impact, resulting in a novel parametric schedule that increased plausible design rates from 83.4% to 93.5% and improved FID from 7.84 to 4.87.
Deep Generative Models (DGMs) are widely used to create innovative designs across multiple industries, ranging from fashion to the automotive sector. In addition to generating images of high visual quality, the task of structural design generation imposes more stringent constrains on the semantic expression, e.g., no floating material or missing part, which we refer to as plausibility in this work. We delve into the impact of noise schedules of diffusion models on the plausibility of the outcome: there exists a range of noise levels at which the model's performance decides the result plausibility. Also, we propose two techniques to determine such a range for a given image set and devise a novel parametric noise schedule for better plausibility. We apply this noise schedule to the training and sampling of the well-known diffusion model EDM and compare it to its default noise schedule. Compared to EDM, our schedule significantly improves the rate of plausible designs from 83.4% to 93.5% and Fréchet Inception Distance (FID) from 7.84 to 4.87. Further applications of advanced image editing tools demonstrate the model's solid understanding of structure.