ppg-in-batch

import numpy as np
import pandas as pd
import os
import torch
from torch.nn import functional as F
from models.deepbeat_bayesian import Bayesian_Deepbeat
import math


def sliding_window(x, w=800, o=400):
    """
    x: 1D numpy array
    w: size of each window
    o: min len of overlapping
    """
    st = 0
    en = 0
    ranges = []
    while en < len(x):
        en = st+w
        if en > len(x):
            en = len(x)
            st = len(x)-w
        ranges.append((st, en))
        # print(st, en)
        st = en-o
    X = np.zeros(shape=(len(ranges),w))
    for r in range(len(ranges)):
        X[r]=x[ranges[r][0]:ranges[r][1]]
    return X


def run_model(model_path, ppg_signals):
    ppg = ppg_signals
    ppg = (ppg-ppg.min())/(ppg.max()-ppg.min())

    batches = sliding_window(ppg, w=800, o=400) #np.zeros((len(ppg)//800, 1, 800))
    batches = batches.reshape(batches.shape[0], 1, batches.shape[1])

    model = Bayesian_Deepbeat()
    model.load_state_dict(torch.load(model_path)['state_dict'])
    model.eval()

    outputs = None
    batches = torch.Tensor(batches)
    with torch.no_grad():
        outputs, _ = model(batches)
        outputs = F.log_softmax(outputs, dim=1)
        outputs = torch.exp(outputs)
    return outputs.numpy()


# reading csv file
# 0-NonAF, 1-AF
def input_preprocessing(model_path, csv_filepath):
    """
    Following program takes csv_directory as input, does necessary operations and then outputs another csv in same directory.
    :param csv_filepath // path where all csv files will be stored
    :return:
    """

    df = pd.read_csv(csv_filepath, engine='python', header=None)
    df = df[df[0]>40]
    ppg = df[1].to_numpy()
    # return ppg
    # return df[1].to_numpy().shape
    return np.argmax(run_model(model_path, ppg), axis=1)

out = input_preprocessing('./bayesbeat_cpu.pt', './uncle_Mar4.csv')

csv_filepath='./uncle_Mar4.csv'
model_path='./bayesbeat_cpu.pt'


# plt.plot(ppg)
# batches.shape
# outputs
# x=np.array([1,2,3,4, 5,6,7,8,9,10,11,12])
# window(x, w=8, o=4)
out