VisiFit: Structuring Iterative Improvement for Novice Designers

Document Title

VisiFit: Structuring Iterative Improvement for Novice Designers

Document Information

• Subject Area: Graphic Design, User Interface Design Tools, Computer-Aided Creativity
• Keywords: Computational Design, Design Tools, Iterative Design, Visual Blending Tools, Novice Designers, User Interface Design, Human-Computer Interaction, Visual Blending

Research Background and Problem

• Identified Problems or Challenges:
  1. Visual blending is an advanced graphic design technique that is particularly challenging for novices, especially when transitioning from low-fidelity prototypes to high-fidelity versions.
  2. Existing tools (e.g., Photoshop) heavily rely on professional expertise and lack features to support iterative improvement for novice designers.
  3. Research shows that while other stages of the design process (e.g., brainstorming, prototyping, evaluation) are well-supported by tools, there is a noticeable lack of tools focused on design iteration.
• Significance:
  1. Iterative design is a critical process for improving design fidelity and achieving target outcomes, but novices currently lack suitable tools to complete this stage effectively.
  2. Providing effective support for novice designers can reduce reliance on experts, foster innovative design creation, and save both time and labor costs.
• Research Motivation and Related Work:
  • Previous tools (e.g., VisiBlends) have supported the generation of initial visual blending prototypes but fail to help users improve the integration of blends.
  • The authors propose a new iterative method, grounded in cognitive science principles of visual perception, to develop tools that assist novices in progressively refining prototypes.

Solution

• Proposed Method or Solution:
  • Developed an iterative improvement method based on secondary design dimensions (color, contours, and internal details).
  • Implemented the VisiFit system, a computational design tool for novice users, offering an interactive and structured workflow.
• Innovations:
  • Introduced the use of secondary design dimensions (e.g., color, contours, internal details) to systematically guide design iteration, leveraging neuroscience principles of human visual object recognition.
  • Implemented a set of interactive, high-level abstraction tools focused on automating secondary design dimension processes (e.g., contour extraction, color blending, and detail extraction).
  • Designed a step-by-step pipeline workflow enabling users to quickly and easily explore the design space.
• Implementation Steps and Key Techniques:
  1. Extract main shapes: Use deep learning and classical computer vision methods (e.g., Grabcut) to segment objects.
  2. Automatic alignment and position adjustment: Adjust object positions via affine transformations.
  3. Select contour options: Provide two automatically generated versions for users to choose the best contour.
  4. Adjust color blending: Implement options such as transparency blending, multiply blending, and color replacement.
  5. Select and reapply internal details: Extract and reposition local details using interactive tools.

Research Outcomes

• Specific Outcomes:
  • Using the VisiFit system, participants significantly improved 76% of initial visual blending prototypes in an average of less than 4 minutes.
  • Among the adjusted designs, 70% achieved a quality suitable for publication on social media.
• Advantages Compared to Existing Solutions:
  • VisiFit provides an intuitive iterative process, avoiding low-level operations (e.g., the complexity of pixel-level manipulation).
  • Novices can quickly create seamless and visually appealing blended designs without requiring professional design expertise.
• Experimental or Evaluation Results:
  • User studies validated the system's effectiveness. Expert reviewers unanimously agreed that the quality of most improved designs was significantly higher than the initial prototypes.
  • The system enables users to explore different secondary dimensions of the design space, making it possible to create more seamless blends.
• Limitations and Future Directions:
  • Limitations: Some detail extraction tools (e.g., reliance on Grabcut) exhibit suboptimal precision in complex scenarios; for certain specific blending problems, the creative outcomes and tool performance remain limited.
  • Future Directions: Expand to more design domains (e.g., animation blending, fashion and furniture design), develop intelligent tools for automatically discovering secondary design dimensions, and further optimize image detail extraction precision.