first commit

benchmark/PaddleOCR_DBNet/utils/metrics.py

@@ -0,0 +1,60 @@
+# Adapted from score written by wkentaro
+# https://github.com/wkentaro/pytorch-fcn/blob/master/torchfcn/utils.py
+
+import numpy as np
+
+
+class runningScore(object):
+    def __init__(self, n_classes):
+        self.n_classes = n_classes
+        self.confusion_matrix = np.zeros((n_classes, n_classes))
+
+    def _fast_hist(self, label_true, label_pred, n_class):
+        mask = (label_true >= 0) & (label_true < n_class)
+
+        if np.sum((label_pred[mask] < 0)) > 0:
+            print(label_pred[label_pred < 0])
+        hist = np.bincount(
+            n_class * label_true[mask].astype(int) + label_pred[mask],
+            minlength=n_class**2,
+        ).reshape(n_class, n_class)
+        return hist
+
+    def update(self, label_trues, label_preds):
+        # print label_trues.dtype, label_preds.dtype
+        for lt, lp in zip(label_trues, label_preds):
+            try:
+                self.confusion_matrix += self._fast_hist(
+                    lt.flatten(), lp.flatten(), self.n_classes
+                )
+            except:
+                pass
+
+    def get_scores(self):
+        """Returns accuracy score evaluation result.
+        - overall accuracy
+        - mean accuracy
+        - mean IU
+        - fwavacc
+        """
+        hist = self.confusion_matrix
+        acc = np.diag(hist).sum() / (hist.sum() + 0.0001)
+        acc_cls = np.diag(hist) / (hist.sum(axis=1) + 0.0001)
+        acc_cls = np.nanmean(acc_cls)
+        iu = np.diag(hist) / (
+            hist.sum(axis=1) + hist.sum(axis=0) - np.diag(hist) + 0.0001
+        )
+        mean_iu = np.nanmean(iu)
+        freq = hist.sum(axis=1) / (hist.sum() + 0.0001)
+        fwavacc = (freq[freq > 0] * iu[freq > 0]).sum()
+        cls_iu = dict(zip(range(self.n_classes), iu))
+
+        return {
+            "Overall Acc": acc,
+            "Mean Acc": acc_cls,
+            "FreqW Acc": fwavacc,
+            "Mean IoU": mean_iu,
+        }, cls_iu
+
+    def reset(self):
+        self.confusion_matrix = np.zeros((self.n_classes, self.n_classes))