Using post label classification and sorting time frames (issues #72 and

#70)

app.py

@@ -17,7 +17,7 @@
 from mmif import Mmif, View, AnnotationTypes, DocumentTypes
 from mmif.utils import video_document_helper as vdh
 
-from modeling import classify, stitch
+from modeling import classify, stitch, negative_label
 
 logging.basicConfig(filename='swt.log', level=logging.DEBUG)
 
@@ -54,23 +54,30 @@ def _annotate(self, mmif: Union[str, dict, Mmif], **parameters) -> Mmif:
         predictions = self.classifier.process_video(vcap)
         timeframes = self.stitcher.create_timeframes(predictions)
 
+        labelset = self.classifier.postbin_labels
+        bins = self.classifier.model_config['bins']
+
         new_view.new_contain(
-            AnnotationTypes.TimeFrame, document=vd.id, timeUnit='milliseconds')
+            AnnotationTypes.TimeFrame,
+            document=vd.id, timeUnit='milliseconds', labelset=labelset)
         new_view.new_contain(
-            AnnotationTypes.TimePoint, document=vd.id, timeUnit='milliseconds')
+            AnnotationTypes.TimePoint,
+            document=vd.id, timeUnit='milliseconds', labelset=labelset)
 
         for tf in timeframes:
             timeframe_annotation = new_view.new_annotation(AnnotationTypes.TimeFrame)
-            timeframe_annotation.add_property("frameType", tf.label),
-            timeframe_annotation.add_property("score", tf.score)
-            timeframe_annotation.add_property("scores", tf.scores)
+            timeframe_annotation.add_property("label", tf.label),
+            timeframe_annotation.add_property('classification', {tf.label: tf.score})
+            #timeframe_annotation.add_property("score", tf.score)
+            #timeframe_annotation.add_property("scores", tf.scores)
             timepoint_annotations = []
             for prediction in tf.targets:
                 timepoint_annotation = new_view.new_annotation(AnnotationTypes.TimePoint)
                 prediction.annotation = timepoint_annotation
                 scores = [prediction.score_for_label(lbl) for lbl in prediction.labels]
                 label = self._label_with_highest_score(prediction.labels, scores)
                 classification = {l:s for l, s in zip(prediction.labels, scores)}
+                classification = self._transform(classification, bins)
                 timepoint_annotation.add_property('timePoint', prediction.timepoint)
                 timepoint_annotation.add_property('label', label)
                 timepoint_annotation.add_property('classification', classification)
@@ -108,6 +115,17 @@ def _label_with_highest_score(labels: list, scores: list) -> str:
         sorted_scores = list(sorted(zip(scores, labels), reverse=True))
         return sorted_scores[0][1]
 
+    @staticmethod
+    def _transform(classification: dict, bins: dict):
+        """Take the raw classification and turn it into a classification of user
+        labels. Also includes modeling.negative_label."""
+        # TODO: this may not work when there is pre-binning
+        transformed = {}
+        for postlabel in bins['post'].keys():
+            score = sum([classification[lbl] for lbl in bins['post'][postlabel]])
+            transformed[postlabel] = score
+        transformed[negative_label] = 1 - sum(transformed.values())
+        return transformed
 
 
 if __name__ == "__main__":

modeling/stitch.py

@@ -47,6 +47,7 @@ def create_timeframes(self, predictions: list) -> list:
         timeframes = self.remove_overlapping_timeframes(timeframes)
         for tf in timeframes:
             tf.set_representatives()
+        timeframes = list(sorted(timeframes, key=(lambda tf: tf.start)))
         if self.debug:
             print_timeframes('Final frames', timeframes)
         return timeframes