Component-Wise Sketching and Generation for Car Interior Design

Paper Title

Component-Wise Sketching and Generation for Car Interior Design

Publication Info

• Topic area: Application of 2D generative AI to car interior design workflows.
• Keywords: 2D generative AI, car interior design, component-wise workflow, 3D cabin template, mixing palette, design exploration, sketching, rendering, user study, professional workflows.

Background and Problem

• Problem / challenge: Current 2D generative AI tools struggle to produce realistic renderings of car interiors from sketches due to the complexity of overlapping components and extreme perspectives. Designers face inefficiencies in exploring and combining numerous component variations manually.
• Significance: Addressing these challenges can streamline car interior design processes, enabling designers to explore more alternatives efficiently and produce high-quality outputs suitable for professional use.
• Motivation and related work: While 2D generative AI has been successfully applied to car exterior design, its application to interiors remains limited. Prior methods often fail to handle complex multi-component scenes or require extensive manual refinement. This paper builds on insights from 3D card-based sketching and aims to extend its applicability to car interiors.

Solution

• Proposed approach: A component-wise workflow for car interior design, where designers sketch individual components, generate refined renderings, compose them into a pre-visualized cabin, and use this as input to generate high-quality cabin renderings.
• Novelty:
  1. Introduction of a component-wise design workflow tailored for car interiors.
  2. Development of the 3D cabin template and mixing palette to support sketching and composing.
  3. Validation of the workflow through a formal user study with professional car designers.
  4. Identification of three distinct design strategies (prepper, cycler, refiner) enabled by the system.
• Procedure and key techniques:
  1. Sketching components: Designers sketch individual components using orthographic and perspective views.
  2. Generating components: 2D generative AI refines these sketches into polished renderings.
  3. Composing cabin: Rendered components are arranged in perspective views using the mixing palette to explore combinations.
  4. Generating cabin: Pre-visualizations are used to produce high-quality cabin renderings.

Results

• Concrete findings:
  • Designers created an average of 3.3 sketches, 6.4 pre-visualizations, and 14.7 cabin renderings per car in 1 hour and 5 minutes.
  • Satisfaction scores: 4.4/5 for usability, 4.6/5 for usefulness, and 8.8/10 for creativity support.
  • Designers demonstrated three strategies: prepper (focused preparation), cycler (iterative exploration), and refiner (incremental refinement).
• Advantage over baselines:
  • Reduced time from weeks to approximately 1 hour per car for generating high-quality cabin renderings.
  • Enabled seamless transitions between sketching, composing, and generating, preserving creative momentum.
  • Supported flexible exploration of design alternatives with minimal manual effort.
• Experiments / evaluation:
  • Participants: 6 professional car designers with an average of 9.2 years of experience.
  • Tasks: Creation of interior concepts for 12 car body types.
  • Metrics: Time spent, number of outputs, satisfaction scores, and creativity support index.
• Limitations and future work:
  • Lack of support for hierarchical modularity and reusable subcomponents.
  • Limited validation outside car interior design; potential applicability to other domains remains unexplored.
  • Future integration with 3D generative AI and VR for immersive design reviews.

Summary

This study introduces a component-wise workflow for car interior design, enabling designers to sketch, generate, and compose individual components into high-quality cabin renderings. The system incorporates a 3D cabin template and mixing palette, facilitating efficient exploration and refinement of design alternatives. A user study with professional designers demonstrated significant time savings, high satisfaction, and applicability to real-world workflows. Future work aims to expand the system's modularity, domain applicability, and integration with 3D generative AI and VR technologies.