Skip to content
/ NeCS Public

Code accompanying the paper "Discovering interaction mechanisms in crowds via deep generative surrogate experiments"

Notifications You must be signed in to change notification settings

kminartz/NeCS

Repository files navigation

Neural Crowd Simulator

This is the code accompanying the paper "Discovering interaction mechanisms in crowds via deep generative surrogate experiments" by Koen Minartz, Fleur Hendriks, Simon Martinus Koop, Alessandro Corbetta, and Vlado Menkovski.

Setup

To install requirements:

pip install -r requirements.txt

Additionally, install a PyTorch version that is compatible with your hardware, and install the corresponding version of torch-scatter. The code was tested with python 3.9.5, pytorch 1.13.1+cu117 and torch-scatter 2.1.1+pt113cu117.

Finally, download the data from this link and unzip it in the root of the repository.

Running the experiments

To run the experiments, run:

python evaluation/run_experiments.py

The calculations can take some time.

Model training

To train your own model, run this command, where <config> is a config file in the configs directory (make sure it is imported in configs/_init_.py, and do not add the .py extension):

python train_model.py <config>

For example:

python train_model.py NeCS_training_default

About

Code accompanying the paper "Discovering interaction mechanisms in crowds via deep generative surrogate experiments"

Resources

Stars

Watchers

Forks

Languages