Classifies Raman spectra using TensorFlow and Keras. This is an old project that is being given a new lease of life so please point any old mistakes / bad practices as an issue.
At the moment, it doesn't even save the model trained, but will hoefully be improved in
git clone https://github.com/rjkilpatrick/raman-cnn/
jupyter notebook raman-cnn
- Python, created in 3.7.4, >= 3.4 should work
- Git, using Gitmoji style commits
- PyTorch
The current implementation, uses a four layer densely connected neural network. The layer sizes were chosen as random powers of two and was ill-informed. In future, this will be moved to a convolutional style neural network.
The accuracy of the model is 90.7%. This is on a small dataset with no data augmentation with a shared training/validation set.
The dataset is taken from https://www.kaggle.com/andriitrelin/cells-raman-spectra/activity, licensed under under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) License.
All code is (c) John Kilpatrick under the MIT license.