C2Ideas: Supporting Creative Interior Color Design Ideation with a Large Language Model

Title of the Paper

C2Ideas: Supporting Creative Interior Color Design Ideation with Large Language Model

Paper Information

• Subject Area: AI-assisted interior color design, particularly creative ideation support based on large language models (LLMs).
• Keywords: Large language model, interior design, color design, human-computer interaction, creative generation, semantic association, interdisciplinary collaboration

Research Background and Issues

• Problems or Challenges:

  • Current interior color design faces issues such as "design generation failing to meet user intent" and "lack of support for design logic."
  • User needs in design are often expressed in vague terms, and existing methods struggle to effectively interpret such abstract requirements.
  • While LLMs possess strong generative capabilities, they lack domain-specific expertise, such as color psychology and principles of color composition.

• Significance:

  • Interior color design directly impacts the aesthetic and emotional effects of a space.
  • Automated color design aligned with design principles and user intent can significantly improve design efficiency and quality.

• Research Motivation and Related Work:

  • Motivation stems from a deep understanding of the limitations of existing AI-assisted design methods:
    • End-to-end generation systems lack transparency and controllability.
    • Insufficient integration of domain knowledge, such as the Natural Color System (NCS).
  • The authors drew inspiration from the successful application of LLMs in other creative fields (e.g., game design and chart design), believing that LLMs also hold potential in interior color design.

Solution

• Method or Solution:

  • A system named C2Ideas is proposed, based on LLMs, which supports interior color design through a three-stage process: Idea Prompting, Word-Color Association, and Interior Coloring.
  • The system integrates domain knowledge and interaction design, leveraging LLM-generated results while allowing multi-level user intervention and adjustments.

• Innovations:

  • Introducing chain-of-thought (CoT) reasoning into color design, making the generation process more interpretable.
  • Embedding domain knowledge (e.g., color psychology, NCS, and color composition principles) into LLM prompt engineering.
  • Providing a flexible user interaction interface, enabling designers to refine design outcomes and enhance control.

• Implementation Steps and Key Techniques:

  1. Idea Prompting (Generating Design Intent):
     • Parsing users' vague requirements into design themes and color emotional attributes.
     • Utilizing carefully crafted few-shot prompt templates, incorporating design context and domain knowledge (e.g., "warm tones convey comfort").
  2. Word-Color Association (Mapping Words to Colors):
     • Translating textual descriptions into specific three-color NCS schemes, considering color proportions and semantic consistency.
     • Integrating natural language processing with color psychology.
  3. Interior Coloring (Spatial Color Assignment):
     • Allocating colors to interior elements based on the color scheme, considering spatial layout and area proportions.
     • Providing 3D scene visualization through virtual reality (VR).
  4. Interactive Design Support:
     • Offering multi-level user interaction, including precise adjustments via graphical interfaces and generalized modifications through natural language.

Research Outcomes

• Specific Results:

  • Developed the C2Ideas system and validated its effectiveness and usability through user studies and expert interviews.
  • The system can generate interior color design schemes that balance user intent, design logic, and domain knowledge.

• Advantages Over Existing Solutions:

  • Easy to interpret and control: The multi-step generation process allows designers to intervene and make adjustments incrementally.
  • High flexibility: Supports various design adjustment methods, from traditional GUI to natural language interaction.
  • Integration of domain knowledge: Incorporates color psychology and NCS principles, resulting in more reasonable and applicable design schemes.

• Experimental or Evaluation Results:

  • User study questionnaires and qualitative interviews indicate that C2Ideas outperforms models lacking domain knowledge in terms of "result relevance," "user preference," rationality, and diversity/inspiration generation.
  • Expert interviews reveal that its step-by-step generation process aligns closely with designers' workflows, and its flexible interaction and strong adaptability are highly praised.

• Limitations and Future Directions:

  • Limitations:
    • Current testing scenarios are limited, and fixed layouts may not fully address complex design needs.
    • More granular design principles and cultural differences in color semantics have not yet been integrated.
  • Future Directions:
    • Expanding scenario adaptability, including dynamic layout adjustments.
    • Providing more customizable color composition paradigms.
    • Balancing the diversity and rationality of generated results, such as supporting personalized designs based on designers' previous work.
    • Enhancing support for image inputs (e.g., mood boards, reference pictures) to improve the system's visual coherence and intuitiveness.

Conclusion

C2Ideas is an innovative system that combines large language models with domain knowledge, offering efficient and targeted support for interior color design through user-centered interaction. It retains creative freedom and allows for manual adjustments, making it a valuable tool for designers.