Aesthify: AI-Powered Design Intelligence

The Business Problem

Design teams assume universal aesthetic principles apply to all audiences. But you can't fit aesthetics into a single curve. What resonates with one demographic fails with another. What works for interior design differs from retail or healthcare spaces.

This creates problems:

Design teams: Apply textbook principles without knowing if they matter to their specific users
Product teams: Run expensive A/B tests instead of asking what their audience actually values
Clients: Get generic design advice when they need audience-specific insights

The opportunity: Build an adaptable pipeline that teams can customize to discover what aesthetic principles matter to their target audience.

The Approach: Customizable Analysis Pipeline

Instead of claiming universal design rules, I built a framework teams can adapt to their context. The process: analyze your designs, quantify aesthetic principles, survey your actual users, discover which principles predict satisfaction for your audience.

Student dorm interior layouts were my proof of concept. I recruited 101 students and showed them 8 room images spanning different timelines, styles, and layouts (from clumsy, cluttered spaces to minimal, clean designs). The pipeline revealed what matters to students choosing dorm aesthetics.

Student Dorm Interior Case Study

The Pipeline in Action

To demonstrate the framework, I focused on student dorm aesthetics:

Step 1: Selected 8 dorm room images deliberately spanning different eras, styles, and layouts (from cluttered, busy spaces to minimal, clean designs)
Step 2: Analyzed images using YOLOv8 and computed scores for 7 aesthetic principles
Step 3: Recruited 101 students to rate how much they'd want each room as their dorm
Step 4: Correlated principle scores with student ratings to discover what predicts dorm preference

What Mattered for Student Dorm Preferences

For student dorm choices, simplicity strongly predicted preference (r=0.68) while symmetry negatively correlated (r=-0.60). Students preferred minimal, uncluttered spaces over formally balanced layouts. This contradicts design textbooks but makes sense for students prioritizing function and calm in shared living spaces.

Key insight: These findings are specific to students evaluating dorm interiors. A different audience (young professionals choosing apartments, families selecting homes) or context (retail stores, hospital waiting rooms) would reveal different priorities. That's the point of the adaptable pipeline.

The Solution: Adaptable Analysis Framework

The system is designed to be customized for different design domains and target audiences. Teams bring their own designs and user base, the pipeline reveals what aesthetic principles matter.

How teams use it:

Input your designs: Upload layouts, mockups, or product photos relevant to your domain
Automated analysis: YOLOv8 detects objects, rule-based scoring quantifies 7 design principles
Survey your users: Collect satisfaction ratings from your actual target audience
Discover insights: Statistical analysis reveals which principles predict preference for your context

Why rule-based scoring? Design principles have established definitions. Rule-based metrics are explainable, auditable, and transferable across domains. Teams can understand exactly how scores are calculated and adapt the rules to their context.

Technical Architecture

Built with Python, Flask, YOLOv8, and Roboflow API. System supports both local (YOLOv8) and cloud (Roboflow) object detection for flexibility.

Seven design principles quantified through computer vision. Each principle uses specific geometric and distributional metrics derived from detected object positions and sizes.

Design Principles Computed:

Balance: Weight distribution across vertical axis using object positions and sizes
Proportion: Golden ratio conformance and object size relationships
Symmetry: Mirror reflection similarity across axes
Simplicity: Object count, density, and visual clutter metrics
Harmony: Color consistency and style cohesion
Contrast: Visual variety in sizes, positions, and object types
Unity: Compositional cohesion and grouping patterns

Technologies: Flask backend, OpenCV for image processing, Pandas for data analysis, Matplotlib for visualizations, Bootstrap for UI.

Business Impact

For Design Teams:

Audience-specific insights: Stop applying generic design rules. Discover what matters to your users.
Evidence-based decisions: Replace "I think this looks better" with data from your target demographic
Faster iteration: Test aesthetic hypotheses without expensive A/B tests

For Product Teams:

Context-specific optimization: Retail store layouts require different principles than healthcare waiting rooms
Demographic segmentation: Understand how preferences vary across age groups, cultures, professions
Design system validation: Test whether your brand aesthetic resonates with intended audiences

Why Adaptability Matters:

No universal aesthetics: Students prioritizing functional dorm spaces value simplicity. Young professionals might value sophistication. Medical interfaces require different principles entirely.
Context shapes preference: Deliberately including cluttered and minimal dorm designs revealed student preferences. Different image sets would reveal different patterns.
Test assumptions: The student dorm study showed symmetry hurts ratings, contradicting design textbooks but making sense for informal living spaces

Real-World Usage

The framework is currently being used by design scholars at IIITDM Kancheepuram for research on aesthetic perception across different cultural contexts and design domains.

Researchers are applying the pipeline to:

Study how Indian vs. Western audiences perceive traditional and contemporary design elements
Analyze which aesthetic principles predict satisfaction in educational spaces
Validate design theories with quantitative evidence from target user populations

This demonstrates the framework's adaptability: same technical infrastructure, different research questions, different insights.

What I Learned

You can't fit aesthetics into a universal curve. The student dorm findings (simplicity matters, symmetry hurts) make sense for students prioritizing functional, calm living spaces. But young professionals might prioritize different principles. A retail store or medical device would reveal completely different priorities. The value isn't in the findings themselves but in the adaptable process.

Diverse image selection is critical. By including dorm layouts spanning different eras and styles (cluttered vs. minimal), I captured the full spectrum of aesthetic variation. This let the analysis reveal genuine preferences rather than artifacts of limited design choices.

Explainability is non-negotiable for creative tools. Rule-based scoring lets teams understand exactly how principles are calculated and adapt them to their domain. A black-box model saying "this design scores 7.2" doesn't help designers improve.

Real-world usage validates the approach. Design scholars at IIITDM are using the framework for cross-cultural studies, showing it works beyond my original student dorm use case. Adaptability was the right design choice.

Next Steps:

Domain-specific modules: Pre-built scoring rules for retail, healthcare, education spaces
Real-time recommendations: Suggest specific changes to optimize for target principles
Integration with design tools: Figma/Sketch plugins for instant feedback during design
Validation studies: Compare predicted preferences vs. actual A/B test results across domains