https://github.com/sorenbouma/keras-oneshot

koch et al, Siamese Networks for one-shot learning, (mostly) reimplimented in keras
https://github.com/sorenbouma/keras-oneshot

Last synced: about 2 months ago
JSON representation

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

• Host: GitHub
• URL: https://github.com/sorenbouma/keras-oneshot
• Owner: sorenbouma
• License: mit
• Created: 2017-03-05T00:02:52.000Z (over 7 years ago)
• Default Branch: master
• Last Pushed: 2019-02-22T02:50:11.000Z (over 5 years ago)
• Last Synced: 2024-02-16T07:37:04.368Z (7 months ago)
• Language: Jupyter Notebook
• Size: 5.45 MB
• Stars: 372
• Watchers: 19
• Forks: 171
• Open Issues: 9
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# keras-oneshot

![oneshot task](images/task_25.png)

[koch et al, Siamese Networks for one-shot learning,](https://www.cs.cmu.edu/~rsalakhu/papers/oneshot1.pdf)  (mostly) reimplimented in keras. 

Trains on the [Omniglot dataset]( https://github.com/brendenlake/omniglot).

Also check out my [blog post](https://sorenbouma.github.io/blog/oneshot) 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

```bash

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.

```bash

git clone https://github.com/brendenlake/omniglot

python load_data.py --path 

```

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.

```bash

jupyter notebook

```