CustomMetrics.md

To use custom metrics, extend easytorch.metrics.ETMetrics the following:

Example for Accuracy, F1 score, Precision, Recall, and IOU:

class Prf1a(ETMetrics):
    r"""
    A class that has GPU based computation of:
        Precision, Recall, F1 Score, Accuracy, and Overlap(IOU).
    """

    def __init__(self):
        super().__init__()
        self.tn, self.fp, self.fn, self.tp = 0, 0, 0, 0

    def add(self, pred, true):
        y_true = true.clone().int().view(1, -1).squeeze()
        y_pred = pred.clone().int().view(1, -1).squeeze()

        y_true[y_true == 255] = 1
        y_pred[y_pred == 255] = 1

        y_true = y_true * 2
        y_cases = y_true + y_pred
        self.tp += _torch.sum(y_cases == 3).item()
        self.fp += _torch.sum(y_cases == 1).item()
        self.tn += _torch.sum(y_cases == 0).item()
        self.fn += _torch.sum(y_cases == 2).item()

    def accumulate(self, other):
        self.tp += other.tp
        self.fp += other.fp
        self.tn += other.tn
        self.fn += other.fn

    def reset(self):
        self.tn, self.fp, self.fn, self.tp = [0] * 4

    @property
    def precision(self):
        p = self.tp / max(self.tp + self.fp, self.eps)
        return round(p, self.num_precision)

    @property
    def recall(self):
        r = self.tp / max(self.tp + self.fn, self.eps)
        return round(r, self.num_precision)

    @property
    def accuracy(self):
        a = (self.tp + self.tn) / \
            max(self.tp + self.fp + self.fn + self.tn, self.eps)
        return round(a, self.num_precision)

    @property
    def f1(self):
        return self.f_beta(beta=1)

    def f_beta(self, beta=1):
        f_beta = (1 + beta ** 2) * self.precision * self.recall / \
                 max(((beta ** 2) * self.precision) + self.recall, self.eps)
        return round(f_beta, self.num_precision)

    def get(self):
        return [self.accuracy, self.f1, self.precision, self.recall]

    def dist_gather(self, device='cpu'):
        serial = _torch.tensor([self.tn, self.fp, self.fn, self.tp]).to(device)
        _dist.all_reduce(serial, op=_dist.ReduceOp.SUM)
        self.tn, self.fp, self.fn, self.tp = serial.cpu().numpy().tolist()