vdplasthijs / PECL

Paired embeddings contrastive learning

Predicting butterfly species presence from satellite data

Resnet-based model to predict species presence vectors from satellite images. The model uses PECL (Paired Embeddings Contrastive Loss) as contrastive regularisation. More details to be added following an upcoming publication.

Installation:

• Use conda to install packages using pecl.yml or pip install from requirements.txt.
• Add your user profile data paths in content/data_paths_pecl.json. (This step is not needed when just experimenting with the code and the example data provided in the repo).

Getting started:

• An example data set is provided in tests/data_tests/.
• Go to notebooks/Getting started.ipynb to see examples of how to load the data and model.
• A link to the full S2-BMS data set will be added soon.

PECL implementation

• Details will follow in the upcoming publication.
• PyTorch implementation can be found in scripts/paired_embeddings_models.py (ImageEncoder.pecl_loss()).
• Models are trained by running scripts/train.py and scripts/train_randomsearch.py.

Results

• The notebooks/ folder contains the notebooks for creating figures/tables.