demo-space2 / neural_methods /loss /NegPearsonLoss.py
swetchareddytukkani
Initial commit with PhysMamba rPPG application
1c6711c
raw
history blame contribute delete
586 Bytes
from __future__ import print_function, division
import torch
import matplotlib.pyplot as plt
import argparse, os
import pandas as pd
import numpy as np
import random
import math
from torchvision import transforms
from torch import nn
class Neg_Pearson(nn.Module):
def __init__(self):
super(Neg_Pearson, self).__init__()
return
def forward(self, preds, labels):
cos = nn.CosineSimilarity(dim=0, eps=1e-6)
pearson = cos(preds - preds.mean(dim=0, keepdim=True), labels - labels.mean(dim=0, keepdim=True))
return torch.mean(1 - pearson)