On or around 4:30pm on Wednesday the 2nd of April, git.uwaterloo.ca will be restarted to apply security and application patches. This maintenance should bring GitLab to patch release 17.10.1
This folder contains code for compiling the [Chest X-Ray demo](https://blogs.technet.microsoft.com/machinelearning/2018/03/07/using-microsoft-ai-to-build-a-lung-disease-prediction-model-using-chest-x-ray-images/) based deep neural network using Athos. [Here](https://github.com/Azure/AzureChestXRay) is the original github repo for the blog, which contains its code.
## Setup
- Run the script `SetupPretrainedTFModel.sh` which sets up the pretrained model in a form which can be subsquently be used by Athos. For this, it first downloads the pretrained Keras model from [this link](https://chestxray.blob.core.windows.net/chestxraytutorial/tutorial_xray/chexray_14_weights_712split_epoch_054_val_loss_191.2588.hdf5)(found from the original github repo). Thereafter, it sets up the [Keras to TensorFlow converter](https://github.com/amir-abdi/keras_to_tensorflow) and converts the Keras based model to the TensorFlow based model and places the same in the folder `./PreTrainedModel/TFModel/`.
- Run the script `SetupPretrainedTFModel.sh` which sets up the pretrained model in a form which can be subsequently used by Athos. For this, it first downloads the pretrained Keras model from [this link](https://chestxray.blob.core.windows.net/chestxraytutorial/tutorial_xray/chexray_14_weights_712split_epoch_054_val_loss_191.2588.hdf5)(found from the original github repo). Thereafter, it sets up the [Keras to TensorFlow converter](https://github.com/amir-abdi/keras_to_tensorflow) and converts the Keras based model to the TensorFlow based model and places the same in the folder `./PreTrainedModel/TFModel/`.
- Once the pretrained model is setup, run the following command to dump the TensorFlow metadata, for further use by Athos.