Exploring the Potentials and Challenges of Deep Generative Models in Product Design Conception

TL;DR

The paper analyzes potentials and barriers for applying deep generative models (DGMs) to early-phase product design conception (PDC), focusing on visual 2D representations to accelerate ideation. It surveys VAEs, GANs, Diffusion models, Transformers, and Radiance Field methods, deriving five key requirements—conditioning/controllability, consistency, coherence, customization, and data/computation costs—for effective integration into PDC. By evaluating each DGM family against these criteria, the work provides model-specific recommendations and highlights practical trade-offs in data, compute, and domain adaptation. The findings indicate diffusion and transformer-based approaches offer the strongest potential for high-fidelity, coherent concept generation, while VAEs and GANs remain valuable for low-to-medium fidelity tasks with greater data efficiency; however, accessibility, domain data, and benchmarks remain critical bottlenecks for industry adoption. The authors propose future research directions in domain-specific datasets, evaluation metrics, and open benchmarks to enable reliable, scalable deployment in engineering design workflows.

Abstract

The synthesis of product design concepts stands at the crux of early-phase development processes for technical products, traditionally posing an intricate interdisciplinary challenge. The application of deep learning methods, particularly Deep Generative Models (DGMs), holds the promise of automating and streamlining manual iterations and therefore introducing heightened levels of innovation and efficiency. However, DGMs have yet to be widely adopted into the synthesis of product design concepts. This paper aims to explore the reasons behind this limited application and derive the requirements for successful integration of these technologies. We systematically analyze DGM-families (VAE, GAN, Diffusion, Transformer, Radiance Field), assessing their strengths, weaknesses, and general applicability for product design conception. Our objective is to provide insights that simplify the decision-making process for engineers, helping them determine which method might be most effective for their specific challenges. Recognizing the rapid evolution of this field, we hope that our analysis contributes to a fundamental understanding and guides practitioners towards the most promising approaches. This work seeks not only to illuminate current challenges but also to propose potential solutions, thereby offering a clear roadmap for leveraging DGMs in the realm of product design conception.

Figures (14)

• Figure 1: Essential components of DGM-driven generation process of Product Design Concepts.
• Figure 2: Examples of visual DGMs not fulfilling the requirements for PDC.
• Figure 3: Interpolation between two classes in the learned CVAE.
• Figure 4: Quantity of papers on ArXiv in the Computer Vision category (cs.CV), mentioning the model type in their abstract
• Figure 5: Quantity of papers on ArX iv in the Machine Learning Category (cs.LG ), mentioning the model type in their abstract.