The official Pytorch code for AutoUnmix, an autoencoder-based spectral unmixing method for multi-color fluorescence microscopy imaging
Run main.py with arguments to train the AutoUnmix on datasets of different fluorophores.
Run fine_tune.py.py to fine-tune AutoUnmix on real biological samples to achieve better reconstruction results.
Run test.py to test the performance of the trained model.
Examples are shown in run.sh.
For the experiments of unmixing performance on simulated datasets, we have provided pretrained models in the repository.
main_simulate.m is used for generate simulated datasets and is modified from LUMoS code.