Trimming boundaries by default

as_seg/data_manipulation.py

@@ -583,14 +583,14 @@ def compute_score_of_segmentation(reference, segments_in_time, window_length = 0
     ref_intervals, useless = mir_eval.util.adjust_intervals(reference,t_min=0)
     est_intervals, useless = mir_eval.util.adjust_intervals(np.array(segments_in_time), t_min=0, t_max=ref_intervals[-1, 1])
     try:
-        return mir_eval.segment.detection(ref_intervals, est_intervals, window = window_length, trim = False)
+        return mir_eval.segment.detection(ref_intervals, est_intervals, window = window_length, trim = True)
     except ValueError:
         cleaned_intervals = []
         #print("A segment is (probably) composed of the same start and end. Can happen with time -> bar -> time conversion, but should'nt happen for data originally segmented in bars.")
         for idx in range(len(est_intervals)):
             if est_intervals[idx][0] != est_intervals[idx][1]:
                 cleaned_intervals.append(est_intervals[idx])
-        return mir_eval.segment.detection(ref_intervals, np.array(cleaned_intervals), window = window_length, trim = False)
+        return mir_eval.segment.detection(ref_intervals, np.array(cleaned_intervals), window = window_length, trim = True)
 
 def compute_median_deviation_of_segmentation(reference, segments_in_time):
     """