The datasets are 1 Hz and 1/3 Hz, so the collection is easily in the range of even cheap processors. On a Rasberry Pi 4 analysis took a couple of minutes, but tests were around 1/2s. It would be interesting to see how it performed on higher sample rates (e.g., PLAID at 30 kHz ot COOLL).
Update: Guess I'm not the only one that thought this might be fun, over 600 entries on github.
Accuracy is a measure of the correct guesses versus all of the guesses
True positive (TP), true negative (TN), false positive (FP), false negative (FN)
The F1 score is a popular performance measure for classification and often preferred when accuracy when
data is unbalanced, such as when the quantity of examples belonging to one class significantly outnumbers
those found in the other class. Precision=TP/(TP+FP) and Recall=TP/(TP+FN)
