• oxnnet_core is a light weight python program designed to be extended to create more complex neural networks for patch based image analysis
• Version 0.01

Create a working directory for the test set OXNNET_TEST_DIR=C:\oxnnet_test_dir `mkdir %OXNNET_TEST_DIR%

Create a set of test cases

python -m tests.utils %OXNNET_TEST_DIR%\TestVolumes

Write out the TensorFlowRecords

python main.py --model oxnnet.model.simplenet write --save_dir %OXNNET_TEST_DIR%\TestRect-tfr --data_dir %OXNNET_TEST_DIR%\TestVolumes\

Train the model

python main.py --model oxnnet.model.simplenet train --tfr_dir %OXNNET_TEST_DIR%\TestRect-tfr --save_dir %OXNNET_TEST_DIR%\TestRect-out --num_epochs 10 --batch_size 100

Test the model (replace <iteration_no> in this command)

python main.py --model oxnnet.model.simplenet test --save_dir %OXNNET_TEST_DIR%\TestRect-out\test --test_data_file %OXNNET_TEST_DIR%\TestRect-tfr/meta_data.txt --model_file %OXNNET_TEST_DIR%\TestRect-out\epoch_model.ckpt-<iteration_no> --batch_size 100

Inspect results Download an NII viewer from https://www.nitrc.org/projects/mricron. Inspect the model outputs in %OXNNET_TEST_DIR%\TestRect-out\val_preds_### and compare them to coresponding test volume.

Dependencies

tflearn (v0.3), tensorflow (v2.3), pandas, nibabel

Python 3.8 recommended.

** How to run tests **

python3 -m unittest