VideoSynchronizationWithPytorch

Code of the paper 'End-To-End Deep Learning Model for Cardiac Cycle Synchronization from Multi-View Angiographic Sequences' (presented at EMBC 2020) using Pytorch.

My python environment : see torch_pip_list.txt

Evaluate a model

1. Open a Jupyter Notebook server with the command jupyter notebook.
2. Open the main.ipynb notebook within the Jupyter Notebook server.
3. Run the following cells:
   • Model declaration
   • Load model (within Test trained model)
   • Load test set (also within Test trained model)
   • Compute distance and similarity matrices for each video
   • Compute distance and similarity matrices for video comparison
   • Pathfinding
   • Global pathfinding result The score will be shown next to the Average total score (Combination) text.

Test on new data

1. Follow the instruction in the README of the AngioMatch repo to prepare the data.
2. Open a Jupyter Notebook server with the command jupyter notebook.
3. Open the main.ipynb notebook within the Jupyter Notebook server.
4. Modify the paths in the "Load test set" cell.
5. Execute the steps needed to evaluate the model.

Train a model on new data

1. Follow the instruction in the README of the AngioMatch repo to prepare the data.
2. Open a Jupyter Notebook server with the command jupyter notebook.
3. Open the main.ipynb notebook within the Jupyter Notebook server.
4. Modify the parameters of the model as you wish and the experiment path in the "Model declaration" cell.
5. Modify the validation_paths and test_paths variable of the "Angio sequence soft multi siamese" cell of the "Load dataset" section.
6. Copy paste the test_paths variable to the "Load test set" cell.
7. Run the following cells:
   • Model declaration
   • Angio sequence soft multi siamese (within Load dataset)
   • Angio sequence multisiamese (within Train)
   • All the ones needed for evaluating a model starting from the third one (Load test set)

Run an hyperparameter search

1. Make sure the right validation and test sets contain the right paths in the load_training_set() method of the hyperparameter_search.py file.
2. If you made changes to the main.ipynb notebook, replicate those changes in the hyperparameter_search.py file.
3. Review the content of the random_parameters variable at the top of the hyperparameter_search.py file.
4. Run the run_hyperparameter_search.bat script.
5. At any time, you can run the read_hyperparameter_search_results.py file to show the results of the hyperparameter search.