Studio272
Announcements
View allAbout
The name Studio272 began with a close group of friends from the advisor’s university days. As an architecture student, he wanted Studio in the name, and 272 was the bus they took home from campus every day.
Today, he carries that warm and familiar name forward as the name of his research group. It also reflects what the group aims to be: a Studio, not a Lab. We aim to go beyond research—to research, design, develop, and build in ways that make a difference in the real world. Also warm research group with warm friends.
This does not mean that we value practical applications above all else. Rather, we believe that strong fundamentals come first. We study the fundamentals deeply, strengthen our understanding step by step, and build upon them carefully. Meaningful applications and innovations grow from a deep understanding of the basics.
Today, building on the advisor’s expertise and experience, Studio272 develops AI-based technologies for the built environment, with a current focus on structural design and construction quality inspection.
- Remote and automated construction quality inspection using computer vision
- Optimization and automation of structural design using optimization and reinforcement learning
But these are only the beginning. The AI technologies we develop can grow into new fields, new industries, and new possibilities. While our current research focuses on the built environment, we welcome problems from any domain. The specific tools may vary, but our work is grounded in the advisor’s expertise in computer vision, optimization, and reinforcement learning. Our next theme could be:
- Smart farming with computer vision, optimization, and reinforcement learning
- Game development with computer vision, optimization, and reinforcement learning
- And beyond—we welcome any topic or domain you want to explore.
We hope you will develop your capabilities, discover your own research or engineering questions, and grow into the researcher or engineer you aspire to become.
Current Research Topics
-
Pix2Pix GAN- and Stereo Vision-Based Automated 3D Reconstruction Without Traditional Feature Matching for Construction Member Dimensional Inspection
-
Remote Dimensional Inspection Using Proportion-Preserving Monocular Perspective Correction and Relief Displacement
-
Vision-Based Real-Time Inclinometer Using SAM2, Sobel, K-Means, Hough, and RANSAC
-
Structural Sizing, Shape, and Topology Optimization via Gradient Descent with a Novel Penalty Function
-
Reinforcement Learning-Based Conceptual Structural Design Optimization
-
Customized Neural Network for Deriving a New Estimation Equation for Concrete Compressive Strength in Existing Buildings
-
Explainability of Deep Learning Models for Strain Distribution Mapping Using Compressed Latent Features
-
Perspective Transformation Using Realistically Similar Images for Building Top-Floor Displacement Measurement
-
Frequency-Domain Image Analysis for Construction Quality Inspection
-
SRGAN-Based Super-Resolution for Construction Member Quality Inspection
-
SinGAN-Based Synthetic Data Generation for Training Construction-Domain AI Models
-
Stereo Vision-Based 3D Reconstruction for Structural Health Monitoring
-
Vision-Based Mathematical Approach to Automated Rebar Counting
