dicarlolab/retinawarp

Tensorflow implementation of fisheye transformation, mimicking the spatial sampling properties in the primate retina

Retina

Tensorflow implementation of fisheye transformation, mimicking the spatial sampling properties of the primate retina

Dependencies

1. numpy

2. scipy

3. scikit-image

4. tensorflow

How to install

1. cd [retina_directory_containing_setup.py]

2. pip install .

How to use

For numpy implementation:

# Import functions
from retina.retina import retina_warp

# read your image
img = imageio.imread('...')
# transform
warp_image(img, output_size=299)

For tensorflow implementation:

# Import functions
from retina.retina_tf import warp_image

# transform
with tf.Session() as sess:
    img = imageio.imread('...')
    retina_img = warp_image(img, output_size=299)
    retina_img = retina_img.eval()

Look here for more details.

Refernce

If you are using this code please refer to our publication:

@article{bashivan2019neural,
  title={Neural population control via deep image synthesis},
  author={Bashivan, Pouya and Kar, Kohitij and DiCarlo, James J},
  journal={Science},
  volume={364},
  number={6439},
  pages={eaav9436},
  year={2019},
  publisher={American Association for the Advancement of Science}
}