Repository for "Evolutionary Generative Design for Defects Shape Reconstruction using Acoustic Measurements" paper submitted to OL2A conference

In the paper, the generative design approach implemented in GEFEST framework was applied to improve optimization process of defect reconstruction problem.

This experiments can execute with Python 3.10.0 enviroment.

• main_experiment.py file includes the code which executes the main experiment mentioned in the paper. This file create Results folder.
• baseline_experiment.py file includes the code which executes the baseline experiment in the paper. This file create Baseline_results folder.
• Paper_results folder contain all results of paper's experiments.
• vizualization folder contain files Historys_visualisation.py(plot a reconstructed shapes and loss of optimization) and evo_results_df.py (create a DataFrame with loss and dice). To vizualize results in this files, string path_to_result can specify the path from the repository root to the experiments.

Installation

To run cases of this repo, need to install GEFEST it's requirements packages.

It's can be installed with pip::

$ git clone https://github.com/ITMO-NSS-team/GEFEST_sound_detection_paper_experiments.git
$ cd GEFEST_sound_detection_paper_experiments
$ pip install -r requirements.txt

Visualisation of optimisation

Other result that not presented in paper

Loss box-plot

Dice box-plot

Proposed approach experiments plots

Baseline plots