THREAD: Exploratory analysis

[gianlucatruda]

This issue can serve as as a thread for exploratory analysis on sample_data/ and nemcova_data

[gianlucatruda]

Starts and ends of sample_data clips may need trimming

I loaded all 11 samples from sample_data/ and compared the mean and std. dev values over time (frames) for all of them. Despite the fact that I already trimmed the starts and ends of the video clips when I created the samples (they looked consistent visually), the summary information shows that the starts and ends are still anomalous. My suggestion here is that when we train on this data, we should perhaps only use frames 500 to 1500 (which corresponds to around 30 seconds).

[MalcolmMielle]

During my (rather small) testing of Nemcova's method, I found that the mean and std of the PPG did vary quite a lot in a given sample thus making it hard to calculate the peaks high and slope. Working on a given section of the signal is one idea but we could also work in smaller incremental windows and then average the results?