DECODE
DECODE copied to clipboard
TuragaLab
This is the official implementation of our publication "Deep learning enables fast and dense single-molecule localization with high accuracy" (Nature Methods)
Some frame processing things modify the frame, after inference emitters are then shifted. Add a method that can modify the emitters accordingly.
Change parameter file logic. Sometimes parameters are exclusive, i.e. if you change LROnPlateau for StepLR the arguments are different and the reference parameter file just adds arguments of StepLR to...
On some Windows machines multiprocessing is unstable (I don't know why this is not reproducible though ...). Maybe we can improve this? Or it automatically improves by upgrading pytorch? Resources:...
Publish a batch fitting script/notebook. There are some open questions: - What do we do if the files are were acquired with different EM gains (--> Write proper tif tag...
Use pytorch-lightning logger implementation which allows for - tensorboard, csv, comet_ml, mlflow etc. - multiple loggers https://pytorch-lightning.readthedocs.io/en/stable/logging.html I think it is perfectly fine to use the logger without anything else...
The stuff that we put in the param.yaml should be also found in the tensorboard logger. Log it there.
X-Y are swapped compared to many other software. Normally that's just because we transpose the frame in the plot function. An easy solution is to remove the transpose and swap...
From @jries: Automatic determination does not work on example data. Why not? Try to fix. Also set auto EM gain as default [in Colab].
INvestigate, rm old stuff and speed up.
Currently when you load params they are autofilled by the reference which is bad style. Seperate