measure fscore besides precision n recall

Former-commit-id: 73c9fcb10a827e5e5a31e8ec2e196327c89db30d

object-locator/eval_precision_recall.py

@@ -41,8 +41,9 @@ def evaluate_sample(self, pts, gt):
     def get_p_n_r(self):
         precision = 100*self.tp / (self.tp + self.fp)
         recall = 100*self.tp / (self.tp + self.fn)
+        fscore = 2 * (precision*recall/(precision+recall))
 
-        return precision, recall
+        return precision, recall, fscore
 
 
 #tests

object-locator/locate.py

@@ -281,19 +281,19 @@
 
     # Output CSV where we will put
     # the precision as a function of r
-    df_prec_n_rec = pd.DataFrame(columns=['precision', 'recall'])
+    df_prec_n_rec = pd.DataFrame(columns=['precision', 'recall', 'fscore'])
     df_prec_n_rec.index.name = 'r'
 
     print(f'\__ Average AHD for all the testing set: {avg_ahd:.3f}')
     print('\__  Accuracy for all the testing set, r=0, ..., 15')
     for judge in judges:
-        prec, rec = judge.get_p_n_r()
-        print(f'r={judge.r} => Precision: {prec:.3f}, Recall: {rec:.3f}')
+        prec, rec, fscore = judge.get_p_n_r()
+        print(f'r={judge.r} => Precision: {prec:.3f}, Recall: {rec:.3f}, F-score: {fscore:.3f}')
 
         # Accumulate precision and recall in the CSV dataframe
-        df = pd.DataFrame(data=[[prec, rec]],
+        df = pd.DataFrame(data=[[prec, rec, fscore]],
                           index=[judge.r],
-                          columns=['precision', 'recall'])
+                          columns=['precision', 'recall', 'fscore'])
         df.index.name = 'r'
         df_prec_n_rec = df_prec_n_rec.append(df)
     print(f'\__  MAPE for all the testing set: {mape:.3f} %')
@@ -309,3 +309,4 @@
 # Write CSV to disk
 df_out.to_csv(os.path.join(args.out_dir, 'estimations.csv'))
 df_prec_n_rec.to_csv(os.path.join(args.out_dir, 'precision_and_recall.csv'))
+