Reimplementation of the paper Sarkar and Etemad 2020 "Self-Supervised ECG Representation Learning for Emotion Recognition"
There exist a Tensorflow 1.14 version from the original author in this repository, however the converters to the datasets are missing
I use anaconda, and the guide assumes you use that too. In case you don't, I guess it is not too hard to modify the install.sh
script to your requirenments.
Anyway, I provided two possible ways to get your conda env working:
conda env create -n conda-env -f sslecg.yml
or you execute the install.sh
script, although you need to manually intervine all the times.
After you did install the environment:
After you did install the environment, there are some more steps to do:
create a folder cache
on the top level, and then one for each dataset you want to use inside, so cache/amigos
,
cache/wesad
and cache/dreamer
.
also edit the file src/constants.py
and change the variables towards the correct
I uploaded the pretrained models, so you can use them like in src/run_example.py
If you want to use the pretraining, you also must download each of the datasets:
All of them must be in a common folder, but you can configure this basepath in src/constants.py