Re-production of SOTA model of CPSC2019.
Original code can be downloaded here.
| test qrs score | train epoch loss |
|---|---|
 |
 |
images can be re-produced via downloading the data csv file from the tfevents file in the results folder, and use MovingAverage in this file to simulate the smooth functionality in the Tensorboard Scalar panel.
NOTE: the ticks of the following plots are unable to be seen in the dark theme!
| test qrs score | train qrs score | train epoch loss |
|---|---|---|
 |
 |
 |
more detailed results can be found here, and the folder results inside of it.
The upper part (with red vertical lines indicating the detected R peaks) is the plot of the result produced by a sophisticated combination of R peak detection algorithms implemented in BioSPPy and wfdb, the lower part (with green vertical lines indicating the detected R peaks) is the plot of the result produced by the original keras models (actually this function). It is clear that the latter outperforms the former.
[1] Cai, Wenjie, and Danqin Hu. "QRS complex detection using novel deep learning neural networks." IEEE Access (2020).