Projects of Ko Yamasaki
文A 2 languages
This article is an outline. It will be expanded only with confirmed details.
Projects of Ko Yamasaki covers independent development, AI systems, coursework, and experimental software created by Ko Yamasaki during university.
The projects vary in scale and completeness, from systems that were actually used to prototypes and learning exercises.
AI systems[edit]
Yamasaki worked on multiple systems using Machine learning.
One project evaluated building design from input images. It was intended not only as a model but as a system through which users could submit images and see inference results.
For the main biographical article, see Ko Yamasaki.
Horse-racing AI[edit]
He also created a horse-racing prediction system using machine learning.
Part of the system was sold, giving him experience with demand for AI-based prediction services. It also raised questions about the limits of models and how expectations around the word "AI" should be handled.
Terraplot[edit]
Terraplot is a location-based game developed in his fourth university year.
Users walk through real places and capture territory enclosed by their movement paths. The project reached beta but was not formally released.
Raspberry Pi and object detection[edit]
In coursework, Yamasaki created an object-detection system using YOLOv5 and a Raspberry Pi vehicle.
The system connected recognition results to physical movement, making it an example of machine learning linked to hardware.
Othello AI[edit]
He created an Othello AI in Lisp.
The implementation used game-tree search and minimax to select moves, providing experience with search algorithms and game AI.
Computer graphics[edit]
In Java computer-graphics work, he implemented Z-buffering, Phong shading, shadows, and related rendering processes.
The work helped him understand how images are generated on screen without relying only on high-level engines.
Generative AI and video[edit]
Later in university, he became interested in AI-assisted media generation.
One experimental system transcribed video audio, stored phrases in a database, searched for phrases close to an input sentence, and reconstructed videos from those fragments.