👀 DeepSeeNet is a deep learning framework for classifying patient-based age-related macular degeneration severity in retinal color fundus photographs.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
data adding source codes Nov 13, 2018
deepseenet fix bugs Dec 7, 2018
docs add source codes Nov 14, 2018
examples fix bugs Dec 7, 2018
images change name Nov 1, 2018
tests adding source codes Nov 14, 2018
.gitignore adding source codes Nov 13, 2018
.travis.yml adding source codes Nov 14, 2018
CONTRIBUTING.md adding source codes Nov 14, 2018
LICENSE.txt adding source codes Nov 14, 2018
README.rst fix bugs Dec 7, 2018
requirements.txt fix bugs Dec 7, 2018



DeepSeeNet is a high-performance deep learning framework for grading of color fundus photographs using the AREDS simplified severity scale. For more details, please see https://ncbi-nlp.github.io/DeepSeeNet/.

Getting Started with DeepSeeNet

These instructions will get you a copy of the project up and run on your local machine for development and testing purposes. The package should successfully install on Linux.



  • python =3.6
  • tensorflow >=1.6.0
  • keras =2.2.4
  • Linux

Tensorflow can be downloaded from https://www.tensorflow.org.

Installing from source

  1. Download the source code from GitHub: git clone https://github.com/ncbi-nlp/DeepSeeNet.git
  2. Change to the directory of DeepSeeNet
  3. Install required packages: pip install -r requirements.txt
  4. Add the code directory to PYTHONPATH: export PYTHONPATH=.:$PYTHONPATH

Using DeepSeeNet for grading simplified scores

The easiest way is to run the following command

$ python examples/predict_simplified_score.py data/left_eye.jpg data/right_eye.jpg
Downloading data from https://github.com/ncbi-nlp/DeepSeeNet/releases/download/0.1/drusen_model.h5
INFO:root:Loading the model: /tmp/.keras/datasets/drusen_model.h5
Downloading data from https://github.com/ncbi-nlp/DeepSeeNet/releases/download/0.1/pigment_model.h5
INFO:root:Loading the model: /tmp/.keras/datasets/pigment_model.h5
Downloading data from https://github.com/ncbi-nlp/DeepSeeNet/releases/download/0.1/advanced_amd_model.h5
INFO:root:Loading the model: /tmp/.keras/datasets/advanced_amd_model.h5
INFO:root:Processing: data/left_eye.jpg
INFO:root:Processing: data/right_eye.jpg
INFO:root:Risk factors: {'pigment': (0, 0), 'advanced_amd': (0, 0), 'drusen': (2, 2)}
The simplified score: 2

The script will

  1. Download the models from the DeepSeeNet repository
  2. Predict the simplified score based on the sample left and right eyes

More options (e.g., setting the models) can be obtained by running

$ python examples/predict_simplified_score.py --help

Pretrained DeepSeeNet models

Besides grading the simplified score, we also provide individual risk factor models. For example

$ python examples/predict_drusen.py data/left_eye.jpg
INFO:root:Loading the model: /tmp/.keras/datasets/drusen_model.h5
INFO:root:Processing: data/left_eye.jpg
The drusen score: [[0.21020733 0.2953384  0.49445423]]
The drusen size: large

All models can be found at deepseenet.

The pretrained models can be found at: https://github.com/ncbi-nlp/DeepSeeNet/releases/tag/0.1

Training DeepSeeNet model

You can train the individual risk factor model too. For example

$ python examples/train.py data/pigment_label_sample.csv data/pigment_best_model.h5
Epoch 1/100
2/2 [==============================] - 27s 14s/step - loss: 1.0103 - acc: 0.5148...
early stopping

The program will read images and labels from a CSV file, train the model, and save the latest best model according to the val_acc.


This work was supported by the Intramural Research Programs of the National Institutes of Health, National Library of Medicine and National Eye Institute.

Citing DeepSeeNet

If you're running the DeepSeeNet framework, please cite:

  • Peng Y, Dharssi S, Chen Q, Keenan T, Agron E, Wong W, Chew E, Lu Z. DeepSeeNet: A deep learning model for automated classification of patientbased age-related macular degeneration severity from color fundus photographs. Ophthalmology. 2018 (Accepted).


This tool shows the results of research conducted in the Computational Biology Branch, NCBI.

The information produced on this website is not intended for direct diagnostic use or medical decision-making without review and oversight by a clinical professional. Individuals should not change their health behavior solely on the basis of information produced on this website. NIH does not independently verify the validity or utility of the information produced by this tool. If you have questions about the information produced on this website, please see a health care professional.

More information about NCBI's disclaimer policy is available.

About text mining group.