A Generative Approach to Light Placement for Street Lighting

This repository contains the source code for the short paper: A Generative Approach to Light Placement for Street Lighting, presented at the Eurographics 2024 conference.

Table of Contents

• Prerequisites
• Downloading the Dataset
• Training the Model
• Testing the Model
• How to Cite
• Acknowledgments
• Contact Us

Prerequisites

The code has been tested on Python 3.9 and is accombanied with a requirements.txt including all dependencies.

Downloading the Dataset

The dataset used in the paper is available here. Download and extract it at the root folder of the project.

To inspect the contents of the dataset navigate to lotus_dataloader.py.

Training the Model

The code is configured to be trained as proposed in the main paper. To instantiate your own training, navigate to main.py script and call the modelTrain() function. Calling the training routine, will automatically create a folder in plots/your_model_name caching all relevant meta data during training.

The file lotus_config.py contains a set of hyperparameters that are used to tune the network. The default parameters are also used in the main paper.

Some extra information:

• You may resume training when creating the model by appropriately setting the argument in model.buildModel(). Please note that this function loads the last recorded model parameters and optimizer's state from meta directory and appends the statistics cached in the plots/your_model_name existing directory

• The meta folder contains the model from the last recorded epoch and a checkpoint of the best recorded model in terms of mean absolute error (see paper) with the name your_model_name_cp.

Testing the Model

To test the trained network there two options. First, on the best recorded checkpoint or second, from a specific checkpoint located in plots/your_model_name/checkpoints. Then, navigate to main.py script and call the modelTest() function. To load a specific model, copy the model cached state to meta/your_model_name and name the file equal to the name variable you used to instantiate modelTest() function. Additionally during testing there are the following options available:

• PLOT_GENERATOR: Plot seeds from the generator
• PLOT_POSTPROCESS: Plot seeds from the generator including the post-process step (see paper)
• EVAL_MODEL: Export several .csv files with detailed metrics per block in the subset at directory plots/your_model_name/evaluation
• EXPORT_LIGHTS: Exports a plots/your_model_name/predicted_lights.json file with the light spatial locations (relative to block dimensions) and their respective intensities

How to Cite

If you use the contents of this repository for your work, please cite it as described below:

In our work, we have used the source code~\cite{Evangelou_EG_2023}, available at ' https://github.com/cgaueb/streetGAN'.

@inproceedings {10.2312:egs.20241031,
	booktitle = {Eurographics 2024 - Short Papers},
	editor = {Hu, Ruizhen and Charalambous, Panayiotis},
	title = {{A Generative Approach to Light Placement for Street Lighting}},
	author = {Evangelou, Iordanis and Vitsas, Nick and Papaioannou, Georgios and Gkaravelis, Anastasios},
	year = {2024},
	publisher = {The Eurographics Association},
	ISSN = {1017-4656},
	ISBN = {978-3-03868-237-0},
	DOI = {10.2312/egs.20241031}
}

Acknowledgments

This research was funded by the Hellenic Foundation for Research and Innovation (HFRI) under the “3rd Call for H.F.R.I. Research Projects to support Post-Doctoral Researchers” (Project No: 7310, LOTUS).

Credits for the datasets:

• Open Street Map

• Open data DC

Contact Us

For any assistance, please contact me at iordanise@aueb.gr. Thank you!