/keras-oneshot

koch et al, Siamese Networks for one-shot learning, (mostly) reimplimented in keras

Primary LanguageJupyter NotebookMIT LicenseMIT

keras-oneshot

oneshot task koch et al, Siamese Networks for one-shot learning, (mostly) reimplimented in keras. Trains on the Omniglot dataset.

Also check out my blog post about this paper and one shot learning in general!

Installation Instructions

To run, you'll first have to clone this repo and install the dependencies

git clone https://github.com/sorenbouma/keras-oneshot
cd keras-oneshot
sudo pip install -r requirements.txt

Then you'll need to download the omniglot dataset and preprocess/pickle it with the load_data.py script.

git clone https://github.com/brendenlake/omniglot
python load_data.py --path <PATH TO THIS FOLDER>

Then you can run the jupyter notebook. If you used python2 to load the data, make sure you use a python2 kernel in your jupyter notebook and vice versa. It's also a good idea to make sure you're using the latest versions of keras and tensorflow.

jupyter notebook