Merge pull request #12 from e1y4r/main

Added kappa and dice score to metrics.

change_detection_pytorch/utils/functional.py

@@ -126,3 +126,56 @@ def recall(pr, gt, eps=1e-7, threshold=None, ignore_channels=None):
     score = (tp + eps) / (tp + fn + eps)
 
     return score
+
+
+def kappa(pr, gt, eps=1e-7, threshold=None, ignore_channels=None):
+    """Calculate kappa score between ground truth and prediction
+    Args:
+        pr (torch.Tensor): A list of predicted elements
+        gt (torch.Tensor):  A list of elements that are to be predicted
+        eps (float): epsilon to avoid zero division
+        threshold: threshold for outputs binarization
+    Returns:
+        float: kappa score
+    """
+
+    pr = _threshold(pr, threshold=threshold)
+    pr, gt = _take_channels(pr, gt, ignore_channels=ignore_channels)
+
+    tp = torch.sum(gt * pr)
+    fp = torch.sum(pr) - tp
+    fn = torch.sum(gt) - tp
+    tn = torch.sum((1 - gt)*(1 - pr))
+
+    N = tp + tn + fp + fn
+    p0 = (tp + tn) / N
+    pe = ((tp + fp) * (tp + fn) + (tn + fp) * (tn + fn)) / (N * N)
+
+    score = (p0 - pe) / (1 - pe)
+
+    return score
+
+
+def dice(pr, gt, eps=1e-7, threshold=None, ignore_channels=None):
+    """Calculate dice score between ground truth and prediction
+    Args:
+        pr (torch.Tensor): A list of predicted elements
+        gt (torch.Tensor):  A list of elements that are to be predicted
+        eps (float): epsilon to avoid zero division
+        threshold: threshold for outputs binarization
+    Returns:
+        float: dice score
+    """
+    pr = _threshold(pr, threshold=threshold)
+    pr, gt = _take_channels(pr, gt, ignore_channels=ignore_channels)
+
+    tp = torch.sum(gt * pr)
+    fp = torch.sum(pr) - tp
+    fn = torch.sum(gt) - tp
+
+    _precision = precision(pr, gt, eps=eps, threshold=threshold, ignore_channels=ignore_channels)
+    _recall = recall(pr, gt, eps=eps, threshold=threshold, ignore_channels=ignore_channels)
+
+    score = 2 * _precision * _recall / (_precision + _recall)
+
+    return score

change_detection_pytorch/utils/metrics.py

@@ -99,4 +99,40 @@ def forward(self, y_pr, y_gt):
         )
 
 
+class Dice(base.Metric):
+
+    def __init__(self,  eps=1e-7, threshold=0.5, activation=None, ignore_channels=None, **kwargs):
+        super().__init__(**kwargs)
+        self.eps = eps
+        self.threshold = threshold
+        self.activation = Activation(activation)
+        self.ignore_channels = ignore_channels
+
+    def forward(self, y_pr, y_gt):
+        y_pr = self.activation(y_pr)
+        return F.dice(
+            y_pr, y_gt,
+            eps=self.eps,
+            threshold=self.threshold,
+            ignore_channels=self.ignore_channels,
+        )
+
+
+class Kappa(base.Metric):
+
+    def __init__(self,  eps=1e-7, threshold=0.5, activation=None, ignore_channels=None, **kwargs):
+        super().__init__(**kwargs)
+        self.eps = eps
+        self.threshold = threshold
+        self.activation = Activation(activation)
+        self.ignore_channels = ignore_channels
+
+    def forward(self, y_pr, y_gt):
+        y_pr = self.activation(y_pr)
+        return F.kappa(
+            y_pr, y_gt,
+            eps=self.eps,
+            threshold=self.threshold,
+            ignore_channels=self.ignore_channels,
+        )