This repository contains the replication material for Estimation of the perceived time of presence of sources in urban acoustic environments using deep learning techniques.
The python folder contains the deep learning model implementation, and the matlab folder contains data processing as well as all other experiments in the paper.
- Clone or download repository from Github.
- Download the experiment corpus from Zenodo and extract its contents to matlab/.
- Download the deep learning dataset from Zenodo and extract its contents to python/data/.
- Install requirements: ''pip3 install -r requirements.txt''
- Run ''presProfileDeep.m'' to generate ground truth presence labels.
- Copy audio files from matlab/audio/rec and matlab/audio/rep to python/data/test_recrep/sound, and from matlab/audio/sim to python/data/test_sim/sound.
- Run ''python3 main.py'' to train the model, compute performance metrics on the evaluation dataset, and generate presence predictions for the perceptual experiment corpus.
- Copy test_recrep_pred.txt and test_sim_pred.txt from python to matlab.
- Run ''paperReplication.m'' to replicate the paper results.
- Felix Gontier ([email protected])
- Mathieu Lagrange ([email protected])