Enterprise AI Analysis
Modeling visual perception of Chinese classical private gardens with image parsing and interpretable machine learning
Abstract: Understanding how spatial visual features shape perception is essential for interpreting the design logic of Chinese classical private gardens, yet perceptual thresholds and the role of spatial organization remain underexplored. This study develops a quantitative framework based on Kaplan's preference matrix, integrating image parsing and interpretable machine learning to examine visual perception in three classical gardens. Visual features, including landscape elements, depth, color, and texture, were evaluated across four perceptual dimensions: coherence, legibility, complexity, and mystery. Results revealed uneven contributions of different features, with nonlinear and threshold effects, while spatial organization shaped perception by regulating the rhythm and intensity of visual features. Waterscapes enhanced coherence, legibility, and mystery. Mountain-view spaces balanced immersion with legibility, and different entrance types either reinforced coherence or encouraged exploration. The framework offers a reproducible approach for linking spatial visual features with perceptual responses in heritage gardens.
Executive Impact: Unlocking Design Insights with AI
This research leverages advanced AI to bridge traditional garden design with modern perceptual science, offering quantitative insights into spatial aesthetics and human experience.
3
Gardens Studied
300
Images Analyzed
35
Visual Features Tracked
4
Perceptual Dimensions
The Challenge: Bridging Tradition & Perception
The intricate design logic of Chinese classical private gardens, with their meticulous control of perceptual rhythm, has largely remained an art form passed down through empirical wisdom. Modern urban integration and scientific evaluation require a deeper understanding of how spatial features truly shape human perception, beyond anecdotal knowledge.
Traditional analyses often oversimplified the complex, multi-dimensional visual features, lacked identification of critical perceptual thresholds, and rarely explored how overall spatial organization influenced perception.
Our AI-Driven Solution: Quantitative Perception Modeling
We developed an innovative framework integrating image parsing and interpretable machine learning with Kaplan's preference matrix. This solution provides a rigorous, quantitative method to analyze how diverse visual features impact perception in heritage gardens.
- Multi-Dimensional Feature Analysis: Expanded visual feature dimensions to 35 indicators covering landscape elements, depth, color, and texture.
- Perceptual Threshold Identification: Identified nonlinear and threshold effects of visual features on coherence, legibility, complexity, and mystery.
- Spatial Organization Impact: Assessed how different spatial strategies regulate visual rhythms and influence perceptual experiences.
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Kaplan's Preference Matrix & Perception
The study deeply explores Kaplan's Preference Matrix, which categorizes human responses to environmental perception into four psychological dimensions: Coherence, Legibility, Complexity, and Mystery. This framework is critical for understanding the cognitive logic behind spatial experiences.
Our analysis reveals that these dimensions are influenced by visual features in a nonlinear fashion, with specific thresholds where perception shifts from positive to neutral or negative. This confirms the nuanced nature of human cognitive processing in complex environments.
Traditional Design Wisdom & Modern Application
Chinese classical private gardens exhibit meticulous control over perceptual rhythm, reflecting centuries of design wisdom. This research translates these traditional principles into quantifiable terms, making them accessible for modern landscape architecture and urban planning.
Insights into how specific elements like waterscapes, mountain views, and entrance types shape coherence, legibility, and mystery provide actionable guidelines for heritage conservation, garden renewal, and the design of new public spaces that aim to evoke similar experiences.
Image Parsing & Interpretable Machine Learning
Leveraging state-of-the-art computer vision techniques, including semantic image segmentation (SegNet) and monocular depth estimation (DINOv2), allowed for the quantitative extraction of 35 visual features from garden images. This overcomes limitations of manual annotation and provides a robust data foundation.
The use of interpretable machine learning models like HGBoost and SHAP enabled the identification of nonlinear relationships and threshold effects, moving beyond simple linear correlations to reveal the complex dynamics between visual input and human perception.
Enterprise Process Flow: Research Workflow
Data Acquisition & Preprocessing
→
Subjective & Objective Evaluation
→
Modeling & Interpretation of Results
35
Visual Features Quantified (Landscape Elements, Depth, Color, Texture) for Comprehensive Analysis.
4
Perceptual Dimensions (Coherence, Legibility, Complexity, Mystery) Analyzed Using Kaplan's Matrix.
Case Study: Optimizing Waterscapes for Perceptual Impact
Problem: Traditional garden practices often employed partial water exposure to encourage exploration, using complex shorelines or barrier plantings, but the precise perceptual impact was anecdotal.
AI-Driven Solution: Quantitative analysis showed that waterscapes achieved peak positive effect on mystery when their visual proportion approached 15%. This aligns with and validates traditional techniques, demonstrating how moderate amounts of natural media divert attention and trigger exploration.
Impact: This strategic placement of waterscapes significantly enhanced coherence, legibility, and mystery in inward-looking spaces, supporting clear spatial orientation and staged cognitive experiences for visitors. Provides a data-backed approach for water feature design.
Case Study: Balancing Immersion in Mountain Views
Problem: Mountain-oriented spaces in historic gardens can sometimes inadvertently create a sense of excessive enclosure, potentially hindering legibility and overall visitor experience.
AI-Driven Solution: The Canglang Pavilion demonstrated an effective strategy for creating open nodes within mountain views by planting low plants (LPVI < 25%), which significantly reduced visual occlusion and provided open vistas. Additionally, ensuring moderate visual cues along paths (RO = 5-8%) supported directional perception.
Impact: This approach successfully balanced immersion with legibility, improving scene coherence and overall visitor flow. Offers actionable insights for designing mountain-view areas to avoid cognitive overload while maintaining an immersive feel.
| Feature | Threshold/Optimal Range | Impact on Perception |
|---|---|---|
| Architectural View Index (AVI) | ~25% (coherence plateau), ~20% (complexity peak) | Coherence: Positive effect stops; Complexity: Peaks then turns negative |
| Road Openness (RO) | ~5% (optimal for coherence/legibility) | Optimal positive effect, then stabilizes |
| Rockery View Index (RVI) | Stabilizes after 30% (negative effect) | Negative effect stabilizes, reduces complexity |
| Saturation Index (STI) | Stops increasing after 60% (positive effect) | Positive effect stops, scene colors muted beyond this |
| Foreground Ratio (FGR) | ~25% (optimal for mystery) | Optimal for mystery, over-masking impedes exploration |
Advanced ROI Calculator
Estimate the potential cost savings and efficiency gains your enterprise could achieve by implementing AI-driven insights from spatial perception research.
Estimated Annual Savings
$0
Hours Reclaimed Annually
0
Your AI Implementation Roadmap
A phased approach to integrating AI-driven spatial perception insights into your enterprise operations.
Phase 01: Discovery & Strategy
Initial consultation to understand your specific challenges and objectives in landscape design, urban planning, or heritage management. Define key performance indicators (KPIs) and scope the AI solution.
Phase 02: Data Integration & Model Training
Collect and integrate relevant visual data (images, plans) from your projects. Customize and train perception models based on your unique spatial contexts and aesthetic goals.
Phase 03: Pilot Project & Validation
Implement the AI framework on a pilot project, analyzing visual features and predicting perceptual outcomes. Validate the model's accuracy against expert evaluations and user feedback.
Phase 04: Full-Scale Deployment & Integration
Scale the AI solution across your entire portfolio or relevant departments. Integrate the AI tool with existing design software or project management platforms for seamless workflow.
Phase 05: Optimization & Ongoing Support
Continuous monitoring, performance tuning, and model refinement based on evolving data and user needs. Provide ongoing technical support and training for your team.
Ready to Transform Your Spatial Design?
Unlock the full potential of AI for understanding and optimizing visual perception in your projects. Schedule a free consultation to see how our solutions can benefit your organization.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat