Chap_3

Classification of retinal data Normal, AMD and DME

Dependencies

Dataset

Prof. Sina Farsiu's team (Duke) and Pratul Srinivasan has generously made the data available here!

Benchmark

Paper

Pretrained model preparation

git clone https://github.com/soumith/inception.torch
cp prepare_model.lua incepion.torch/
cd incepion.toch
th prepare_model.lua

This creates inception.t7 in working directory

Data preparation + fine-tuning+ testing

1_main_train: Preprocess training data.

1_main_test: Preprocess testing data.

1_main_train_unprocess: Only resizing of train data.

1_main_test_unprocess: Only resizing of test data.

2_data.py: Creates labels corresponding to traina dn test data, randomizes train data, Compute mean image. and saves in a hdf5 file.

3_main_auxi: Load pretrained GoogLenet, fine-tune, decision pooling and confidence on test set.

4_ main_psuedo_error: identified indexes of representative response at each layer and saves in error folder.

5_ representative_responses: vizualization of representative responses.

Cross-validation

https://github.com/ultrai/Chap_3/blob/master/main_cuda_proper_cv.lua#L55-L58

Class weights	loss weights	Decision pooling accuracy
0.3 0.3 0.4	1 1 1	0.99 0.89 0.84
0.3 0.3 0.4	1 0.1 0.0001	0.99 0.89 0.86
0.2 0.2 0.6	1 1.00E-07 1.00E-14	0.98 0.85 0.81
0.33 0.33 0.33	1 1 1	1 0.88 0.81
0.33 0.33 0.33	1 0.0 0.0	0.98 0.85 0.85

The MIT License (MIT)

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.