Dynamice beta Variational Auto Encoder (VAE) for biodiversity assessment of insect signals. A fully unsupervised model is outperforming conventional methods, such as PCA whereas a semi-supervised method improves upon the unsupervised model results even further.
This code is made publicly available together with the article Dynamic beta VAEs for quantifying biodiversity by clustering optically recorded insect signals, Klas Rydhmer and Raghavendra Selvan, accepted for publication by Ecological Informatics, 2021-10-05.
This repository provides a minimum working example of the code. As the insect signals used in the published work are used commercially by FaunaPhotonics, they are not included in this repository. Instead, a framework for generating synthesized signals is provided.
-
Basic Pytorch dependency pip install -r requirements.txt
-
Tested on Pytorch 1.3, Python 3.6
-
Synthesize new sample data:
python Synthesize_signals.py -
Train the model from scratch: python DynamicBetaVAE.py
- Kindly cite our publication if you use any part of the code
@article{rydhmer2021dynamicVAE,
title={Dynamic beta VAEs for quantifying biodiversity by clustering optically recorded insect signals},
author={Klas Rydhmer and Raghavendra Selvan},
journal={Ecological Informatics},
month={October},
note={arXiv preprint arXiv:2102.05526},
year={2021}}