End-to-End Automatic Morphological Classification of Intracranial Pressure Pulse Waveforms Using Deep Learning
To create anaconda environment for this project you need to open Anaconda Prompt, change working directory to project folder and run the script
conda env create -f environment.yml
[.] - folder
--[datasets]
|--[full_dataset]
| |--[train]
| | |-files in the form Class_ID.csv
| |--[test]
| | |-files in the form Class_ID.csv
|--[RAW_dataset]
| |-provided dataset
|--[RAW_corrected_dataset]
| |-provided dataset
--[experiments]
|--[ModelName_ID]
| |--[model_weights]
| | |-model_final.pth
| | |-model_best.pth
| |-tensorboard events file
--[models]
|-Files with code for the models
--[signal_processing]
|-Files with code for the initial processing of the ICP/ABP signals
-utils.py:
|-Dataset loader class, other utility functions
-training_loop.py:
|-Training loop with tensorboard hooks and saver
-experiment_manager.py:
|-Experiment management class, example of learning the network
-multiexp.py
|-Multiple experiments management structure
-reformat_dataset.py
|-Script for transforming raw dataset into training and testing datasets with unique id's and mapping
You need raw dataset in the datasets folder. Modify the code in the reformat dataset to suit your needs - best leave new dataset as full_corrected_dataset Then run:
python reformat_dataset.py
python experiment_manager.py
or for batch learning
python multiexp.py