Training Climbing Roses By Constrained Graph Search

Computer Graphics International 2024

Authors: Wataru Umezawa and Tomohiko Mukai

Download: [preprint] [slides]

Abstract: Cultivated climbing roses are skillfully shaped by arranging their stems manually against support walls to enhance their aesthetic appeal. This study introduces a procedural technique designed to replicate the branching pattern of climbing roses, simulating the manual training process. The central idea of the proposed approach is the conceptualization of tree modeling as a constrained path-finding problem. The primary goal is to optimize the stem structure to achieve the most impressive floral display. The proposed method operates iteratively, generating multiple stems while applying the objective function in each iteration for maximizing coverage on the support wall. Our approach offers a diverse range of tree forms employing only a few parameters, which eliminates the requirement for specialized knowledge in cultivation or plant ecology.

Acknowledgement: The authors thank Takunori Kimura of Rosa Orientis gave us very useful feedback on the resulting images from the perspective of an expert rose breeder. Japan Rose Society also provided us with many suggestions for improvement through evaluation experiments on the generated images. This work was supported by JST SPRING, Grant Number JPMJSP2156 and PlatinumGames Inc.

1 thought on “Training Climbing Roses By Constrained Graph Search

  1. Pingback: Two Papers at Computer Graphics International 2024 – Mukai Laboratory

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.