Colorbo: Envisioned Mandala Coloring through Human-AI Collaboration

Title of the Paper

Colorbo: Envisioned Mandala Coloring through Human-AI Collaboration

Paper Information

• Domain: Human-AI Collaborative Design, Computer-Aided Art Creation
• Keywords: Coloring, Mandala, Human-AI Collaboration, Interactive System, Artificial Intelligence, Visual Projection, Mental Health, Artistic Creation, Color Recommendation, User Experience

Research Background and Problem

• Problems and Challenges:
  • Mandala coloring, as a popular activity, can enhance focus, reduce anxiety, and improve mental health. However, when coloring mandalas manually on paper, users often face challenges:
    • How to select harmonious color combinations
    • Difficulty in predicting the final effect after color filling
  • These challenges make it harder for users to achieve their intended goals, leading to a gap between the finished product and their initial vision.
• Significance:
  • Mandala coloring is widely used in psychotherapy and stress relief. Addressing user experience pain points can improve satisfaction with the activity and promote its broader adoption.
• Research Motivation:
  • Current AI interactions in artistic creation are mostly based on improvisational collaboration, lacking methods to help users construct a "mental image" of their work.
  • This study aims to provide users with visual suggestions through human-AI collaboration, helping them "envision" the potential development of their creations.

Solution

• The authors propose an interactive system called Colorbo:
  • System Logic:
    • Colorbo and the user color the same mandala sketch simultaneously, with real-time observation of each other's progress.
    • The user displays a partially colored sketch, and Colorbo analyzes the remaining patterns and the user's color combinations to automatically complete the coloring.
    • The fully colored sketch is projected onto the user's paper using a projector, helping them continue coloring and "envision" the final effect.
  • Technical Innovations:
    • A deep learning-based color embedding model: Automatically generates colors harmonious with those already used by the user.
    • Offers two visual projection modes:
      1. Full Image Mode: Projects the entire image.
      2. Layered Mode: Projects in layers, emphasizing specific areas.
  • Implementation Steps:
    1. Capture and preprocess the user's current work image, including edge cropping, image alignment, and color quantization.
    2. Segment the sketch into regions and group them into segments.
    3. Use the color embedding model to select colors that coordinate with the user's work.
    4. Color the remaining areas, generating multiple complete sketch versions for the user to choose from.
    5. Project the sample sketches generated by Colorbo onto the user's coloring paper using a projector.

Research Outcomes

Specific Achievements

• Positive Experience:
  • Colorbo effectively improved user satisfaction with their final works.
  • The system helped users construct a "mental image," predict future results, and form a creative plan.
  • Reduced user stress in color selection, avoiding anxiety caused by decision-making difficulties.
• Beneficial Features:
  • The system-generated color combinations were considered harmonious and aesthetically pleasing.
  • Users enjoyed the diverse and unexpected color suggestions provided by the system.

Advantages Compared to Existing Solutions

• Introduced a novel AI collaboration approach based on "envisioning" rather than traditional improvisational support.
• The projection-based presentation enhanced the interactive experience of paper-based coloring, addressing the limitations of manual coloring, such as difficulty in making changes.

Experimental or Evaluation Results

1. In a controlled experiment with 16 participants (university students), subjects colored two different mandalas using Colorbo and traditional methods, respectively.
   • Satisfaction with the works was significantly higher when using Colorbo.
   • Although coloring time slightly increased, users felt the system more effectively helped them achieve their desired creative outcomes.
2. Comparison of Projection Modes:
   • Over 62.5% of users preferred the Full Image Mode, as it provided a global view for overall planning.
   • Users focused on details showed a preference for the Layered Mode.
3. The system-generated color combinations achieved user satisfaction comparable to popular color schemes.

Limitations and Future Directions

• Current Limitations:
  • The study tested only fixed sketches and coloring tools; future research should validate the system's adaptability to different materials and complexities.
  • The matching between projected colors and actual colors needs further optimization.
  • Participants were all students, who may be more familiar with technology; broader user groups need to be included.
• Future Directions:
  • Explore more "envisioning modes" tailored to different user needs.
  • Develop dynamic and evolving intelligent collaboration systems to prevent over-reliance on Colorbo that might hinder creativity.
  • Extend the project to other art forms to verify its generalizability.

This study presents a novel collaboration model between artificial intelligence and humans in the field of artistic creation, proposing an innovative "envisioning" approach. It contributes to the creation of more harmonious and aesthetically pleasing artworks and broadens the research scope of human-computer interaction.