fix: stop FEs incorrectly dropping some Features

_keep_num() was incorrectly detecting overlaps

pipeline/feature_extractors.py

@@ -254,14 +254,14 @@ class LoudAudioFeatureExtractor(FeatureExtractor):
 
         return loudness_features
 
-    def _keep_num(self, features, num=_CONFIG_DEFAULT_NUM_FEATURES, margin=10.0) -> list:
+    def _keep_num(self, features, num=_CONFIG_DEFAULT_NUM_FEATURES, trim_overlap=False) -> list:
         """Keep the top n features (default: 5)
 
         Approach:
          - for range in 0-n
            + expand the nth top feature to min duration
             (move start back by 0.5*min_duration, end forward by 0.5*min_duration)
-           + drop any features that are now in that feature's range (plus margin)
+           + drop any features that are now in that feature's range (optional)
          - return the top n features
 
         Each feature is a Feature object, with an Interval object
@@ -271,14 +271,20 @@ class LoudAudioFeatureExtractor(FeatureExtractor):
         features = sorted(features, key=lambda x: x.score, reverse=True)
         for i in range(num):
             current_feature = features.pop(0)
-            # expand the feature to min_duration
-            current_feature.interval.move_start(-0.5*self._min_duration, relative=True)
-            current_feature.interval.move_end(0.5*self._min_duration, relative=True)
+            # expand the feature to min_duration - try and keep centered at current start
+            if self._min_duration > current_feature.interval.duration:
+                current_feature.interval.move_start(-0.5*self._min_duration, relative=True)
+            if current_feature.interval.duration < self._min_duration:
+                current_feature.interval.update_duration(self._min_duration)
             keep_features.append(current_feature)
             # drop any features that are now in that feature's range (plus margin)
-            features = [f for f in features if
-                        (f.interval.end < current_feature.interval.start-margin or
-                         f.interval.start > current_feature.interval.end+margin)]
+            # features = [f for f in features if
+            #             (f.interval.start < current_feature.interval.start-margin and
+            #             f.interval.end > current_feature.interval.start-margin) or
+            #              (f.interval.end > current_feature.interval.end+margin and
+            #               f.interval.start < current_feature.interval.end+margin)]
+            if trim_overlap:
+                features = [f for f in features if f.interval.overlaps(current_feature.interval)]
 
         return keep_features
 
@@ -305,8 +311,12 @@ class LoudAudioFeatureExtractor(FeatureExtractor):
                 features.append(Feature(interval=Interval(start=time, duration=0.500),
                                              source=file, feature_extractor="loudness",
                                              score=loudness))
+
             # prune features list to keep self.num_features
-            self.features = self._keep_num(features, self._num_features)
+            if len(features) > self._num_features:
+                self.features = self._keep_num(features, self._num_features)
+            else:
+                self.features = features
 
 
 class VideoActivityFeatureExtractor(FeatureExtractor):
@@ -383,14 +393,14 @@ class VideoActivityFeatureExtractor(FeatureExtractor):
         scores = sorted(scores, key=lambda x: x[1], reverse=True)
         return scores[:int(len(scores) * (percent / 100))]
 
-    def _keep_num(self, features, num=_CONFIG_DEFAULT_NUM_FEATURES, margin=10.0) -> list:
+    def _keep_num(self, features, num=_CONFIG_DEFAULT_NUM_FEATURES, trim_overlap=False) -> list:
         """Keep the top n features (default: 5)
 
         Approach:
          - for range in 0-n
            + expand the nth top feature to min duration
             (move start back by 0.5*min_duration, end forward by 0.5*min_duration)
-           + drop any features that are now in that feature's range (plus margin)
+           + drop any features that are now in that feature's range (optional)
          - return the top n features
 
         Each feature is a Feature object, with an Interval object
@@ -400,18 +410,23 @@ class VideoActivityFeatureExtractor(FeatureExtractor):
         features = sorted(features, key=lambda x: x.score, reverse=True)
         for i in range(num):
             current_feature = features.pop(0)
-            # expand the feature to min_duration
-            current_feature.interval.move_start(-0.5*self._min_duration, relative=True)
-            current_feature.interval.move_end(0.5*self._min_duration, relative=True)
+            # expand the feature to min_duration - try and keep centered at current start
+            if self._min_duration > current_feature.interval.duration:
+                current_feature.interval.move_start(-0.5*self._min_duration, relative=True)
+            if current_feature.interval.duration < self._min_duration:
+                current_feature.interval.update_duration(self._min_duration)
             keep_features.append(current_feature)
             # drop any features that are now in that feature's range (plus margin)
-            features = [f for f in features if
-                        (f.interval.end < current_feature.interval.start-margin or
-                         f.interval.start > current_feature.interval.end+margin)]
+            # features = [f for f in features if
+            #             (f.interval.start < current_feature.interval.start-margin and
+            #             f.interval.end > current_feature.interval.start-margin) or
+            #              (f.interval.end > current_feature.interval.end+margin and
+            #               f.interval.start < current_feature.interval.end+margin)]
+            if trim_overlap:
+                features = [f for f in features if f.interval.overlaps(current_feature.interval)]
 
         return keep_features
 
-
     def setup(self):
         pass
 
@@ -427,7 +442,10 @@ class VideoActivityFeatureExtractor(FeatureExtractor):
                                              score=score))
 
             # prune features list to keep self.num_features
-            self.features = self._keep_num(features, self._num_features)
+            if len(self.features) > self._num_features:
+                self.features = self._keep_num(features, self._num_features)
+            else:
+                self.features = features
 
     def teardown(self):
         pass