Yolo_Apnea_Predicter

NB! Still a work in progress, expect interface to change

Master thesis code for predicting apnea events using yolo. Predict Apnea events on ABDO signal from .edf file or other numpy array signal using trainled Yolo model. Outputs nsrr-xml info, but wil later be expanded to be able to compare to true signals annotated by sleep technicians, and will return values of how good the predictions are.

When more models have been generated and trained, the intention is that this repo will handle them as well by changing paramenters when initializing the detector

Description

Predict Apnea events on .edf file or ABDO_RES signal and returns an predictions object with different ways to visualize the predictions.

This package can be run as a standalone detector, or can be imported in other projects

Imported in other projects

Usage:

Run yolo on signal:

Import as:

from yoloapnea.apnea_detector import ApneaDetector

Instantiate:

detector = ApneaDetector()

Run yolo on signal:

signal = np.array of shape (x,) # Currently needs to be ABDO_RES signal
detector.append_signal(signal)

Access predictions

Needs to be done after the previous part to have generated predictions.

Get prediction object:

predictions = detector.predictions

Get NSRR xml output of predictions:

xml = predictions.get_xml()

Get numpy array of predictions

xml = predictions.get_predictions_as_np_array()

Get numpy array of the most recent predictions Usefull for real-time detection. Returns an array of the same length as the sliding window durtation

xml = predictions.get_last_predictions()

---

Standalone from main

Usage:

usage: main.py [-h] file

Arguments

short	long	default	help
-h	--help		show this help message and exit