An interactive web interface for exploring synthetic one-bedroom (1+1) apartment layout generation through controllable constraints, distributions, and spatial rules.
This project accompanies and extends an academic research effort on using synthetic data as a design and training medium for architectural intelligence systems.
The Synthetic Layout Generator UI allows users to interactively generate and inspect one-bedroom apartment (1+1) layouts produced by a rule-based synthetic generator.
Rather than producing a single “optimal” plan, the system focuses on:
- Variation
- Constraint logic
- Adjacency rules
- Cultural and regulatory assumptions embedded in layouts
The UI acts as a bridge between research logic and spatial intuition.
This is not a production floor-plan generator.
It is an exploratory, analytical, and educational tool.
This interface is derived from the master’s thesis:
“Training GANs with Synthetic Data: A Dual-Layered Approach to AI-Driven Architectural Layout Generation”
Mehmet Sadık Aksu, 2025
The core research investigates:
- How architectural layouts can be synthetically generated using rule-based systems
- How such datasets can be used to train generative models
- How spatial, cultural, and regulatory biases appear in generated data
To maintain a focused and controlled scope, the research — and this UI — specifically targets
one-bedroom apartment units (1+1 flats) as a foundational residential typology.
The UI was developed to:
- Externalize the generator’s assumptions
- Make synthetic data generation inspectable and adjustable
- Support research communication and experimentation
📄 Related blog post:
https://sadikaksu.com/blog/synthetic-layout-generator
Architectural generators are often hidden behind code and parameters.
This UI was built to:
- Make design logic visible
- Allow non-programmers to explore generative systems
- Enable rapid testing of parameter ranges
- Serve as a research demonstrator rather than a black box
- Generate synthetic one-bedroom apartment (1+1) layouts in real time
- Control global parameters such as:
- Unit size ranges
- Room distributions
- Adjacency constraints
- Visualize layouts as simplified architectural diagrams
- Export and inspect generated results
- Observe how small parameter changes affect spatial outcomes
-
Open the live demo
👉 https://sadikaksu.github.io/synthetic-layout-ui/ -
Adjust parameters using the control panel
-
Regenerate layouts to explore variation
-
Observe recurring spatial patterns and anomalies within the one-bedroom unit typology
synthetic-layout-ui/
├── index.html
├── src/
│ ├── generator/
│ ├── ui/
│ └── utils/
├── docs/
│ └── images/
├── LICENSE.md
└── README.md
This project is licensed under Creative Commons Attribution–NonCommercial (CC BY-NC 4.0).
You are free to:
- Share and adapt the work
- Use it for research, education, and non-commercial purposes
Under the following conditions:
- Attribution is required
- Commercial use is not permitted
See LICENSE.md for details.
If you use this project in academic or research work, please cite:
Aksu, M. S. (2025). Training GANs with Synthetic Data: A Dual-Layered Approach to AI-Driven Architectural Layout Generation.
Developed as part of ongoing research at the intersection of:
- Architecture
- Generative systems
- Artificial intelligence
- Synthetic data methodologies