Automatic Creation Of 3D Art Using Multi-Objective Creative Evolutionary Design Algorithms
Authors: Lee Chin Kuan and Jason Teo Tze Wi
In this research, we proposed Genetic Algorithms (GA) with Multiple Evaluation as the evolutionary algorithm to optimize multi-objective evaluation functions in 3D art. GA is chosen to evolve the parameters of the 3D art for the Gieles Superformula’s m1 parameter. Several evaluations using the multi-objective fitness functions are taken to evaluate the symmetry and radius differential of the 3D art. Four types of experiment have been conducted to test which type of combination of crossover rate and mutation rate produced the best individual among each other. These experiments are: Experiment 1: 50 percent crossover and 1 percent mutation; Experiment 2: 50 percent crossover and 10 percent mutation; Experiment 3: 90 percent crossover and 1 percent mutation; and Experiment 4: 90 percent crossover and 10 percent mutation. Findings are that from Experiment 2, which is 50 percent crossover and 10 percent mutation, performed the best among all the experiments. There were a total of 6 out of 9 individuals that lies on the Pareto curve of the overall experiment. User evaluations were then conducted to verify whether or not the shapes after evolution looks more pleasing than before the shapes were evolved. There were a total of 9 shapes that were shown to the human evaluators. 6 of them were shown on the laptop screen, 3 of them were shown by 3D printed model. 4 out of 6 of the shapes shown on the laptop screen were chosen from each experiment’s final product, while the other 2 were unevolved objects. 1 out of 3 3D printed model was the evolved product of experiment 2, whereas the other 2 were unevolved objects. The user evaluations showed that overall users preferred the product after evolution compared to the object before being evolved. Therefore, this research has shown that the implementation of a Multi-Objective Evolutionary Algorithm in optimizing 3D art is able to create abstract 3D objects which appear to be aesthetically pleasing to human evaluators.