TextureAda: Deep 3D texture transfer for ideation in product design conceptualization

In the product design life-cycle, the conceptual design stage is an important and time-consuming phase where designers ideate in the form of sketches and 3D renderings of a product. Specifically, with 3D renderings, the choice of material texture and color is an important aspect that is often critiqued by designers because it impacts the product’s visual aesthetic and the impression it evokes in the customer when first viewing the product; thus, making material selection and the conceptual design stage, overall, challenging. In this study, we turn to deep texture synthesis for generating material textures and propose a novel method, TextureAda. TextureAda creates high-fidelity textures by performing adaptive instance normalization between multiple layers of a texture generator and a pre-trained image encoder. Our experiments show that our method beats previous methods in texture synthesis visually and quantitatively. Lastly, we show how TextureAda can be applied for ideation in product design conceptualization by material texturing 3D models of furniture.