2023-08-30 17:43:59 +02:00
|
|
|
# Machine learning for material recognition with a triboelectric nanogenerator (TENG)
|
|
|
|
This project was written for my bachelor's thesis.
|
|
|
|
|
|
|
|
It was written to classify TENG voltage output from pressing it against different materials.
|
|
|
|
Contents:
|
|
|
|
- Data preparation/plotting/loading utilites
|
|
|
|
- (Bi)LSTM + fully connected + softmax model for name classifiying TENG output
|
|
|
|
- Progress tracking utilities to easily find the best parameters
|
2023-05-10 22:44:14 +02:00
|
|
|
|
|
|
|
## Model training
|
2023-08-30 17:43:59 +02:00
|
|
|
Adjust the parameters in `main.py` and run it.
|
2023-05-10 22:44:14 +02:00
|
|
|
All models and the settings they were trained with are automatically serialized with pickle and stored in a subfolder
|
|
|
|
of the `<model_dir>` that was set in `main.py`.
|
|
|
|
|
|
|
|
|
|
|
|
## Model evaluation
|
|
|
|
Run `find_best_model.py <model_dir>` with the `<model_dir>` specified in `main.py` during training.
|