Awesome

Variational Autoencoder with Gaussian Random Field prior

Repository linked with the publication

Variational Autoencoder with Gaussian Random Field prior: application to unsupervised animal detection in aerial images, H. Gangloff, M.-T. Pham, L. Courtrai, S. Lefèvre, 2022. (https://hal.archives-ouvertes.fr/view/index/docid/3774853)

The model can be directly tested on the Livestock dataset which is provided to reproduce the results from this section of the article.

To train a model run the file: sh vae_train.sh. For the classical VAE model, set corr_type=corr_id, for the VAE-GRF model set corr_type=corr_exp or corr_type=corr_m32. Dataset available is livestock for now.

To test a model run the file: sh vae_test.sh with appropriate parameters. Some checkpoints files are provided in torch_checkpoints to reproduce directly the results from the article.

The code is built with PyTorch and other standard librairies.

For more details, refer to the publication.