Merge pull request #51 from KumarLabJax/distance-filter-experiments

Adding extra features to run comparisons

Author: SkepticRaven

src/jabs_postprocess/heuristic_classifiers/locomotion_corner.yaml

@@ -0,0 +1,43 @@
+# Locomotion in corner
+# Within 20% distance of any 2 walls
+# Mouse is moving > 5cm/s
+# This is really slow for a mouse, but essentially just removes noise from micromovements
+behavior: 
+  Locomotion In Corner
+
+interpolate:
+  5
+
+stitch:
+  5
+
+min_bout:
+  15
+
+definition:
+  and:
+    - minimum:
+      - 2
+      - less than:
+        - wall_distances/wall_0
+        - divide:
+          - avg_wall_length
+          - 5
+      - less than:
+        - wall_distances/wall_1
+        - divide:
+          - avg_wall_length
+          - 5
+      - less than:
+        - wall_distances/wall_2
+        - divide:
+          - avg_wall_length
+          - 5
+      - less than:
+        - wall_distances/wall_3
+        - divide:
+          - avg_wall_length
+          - 5
+    - greater than:
+      - features/per_frame/centroid_velocity_mag centroid_velocity_mag
+      - 5.0

src/jabs_postprocess/heuristic_classifiers/locomotion_periphery.yaml

@@ -0,0 +1,26 @@
+# Locomotion in periphery
+# Within 20% distance of any wall
+# Mouse is moving > 5cm/s
+# This is really slow for a mouse, but essentially just removes noise from micromovements
+behavior: 
+  Locomotion In Periphery
+
+interpolate:
+  5
+
+stitch:
+  5
+
+min_bout:
+  15
+
+definition:
+  and:
+    - less than:
+      - features/per_frame/corner_distances distance to wall
+      - divide:
+        - avg_wall_length
+        - 5
+    - greater than:
+      - features/per_frame/centroid_velocity_mag centroid_velocity_mag
+      - 5.0

src/jabs_postprocess/utils/project_utils.py

@@ -509,6 +509,8 @@ def __init__(self, settings: ClassifierSettings, data: pd.DataFrame):
             "exp_prefix",
             "time",
             "distance",
+            "distance_threshold",
+            "distance_seg",
             "closest_id",
             "closest_lixit",
             "closest_corner",
@@ -677,6 +679,8 @@ def __init__(self, settings: ClassifierSettings, data: pd.DataFrame):
             "exp_prefix",
             "time",
             "distance",
+            "distance_threshold",
+            "distance_seg",
             "closest_id",
             "closest_lixit",
             "closest_corner",
@@ -770,6 +774,16 @@ def add_bout_features(self, feature_file: Path):
                 "features/per_frame/centroid_velocity_mag centroid_velocity_mag",
                 lambda x: np.nansum(x, initial=0) / 30,
             ),
+            FeatureInEvent(
+                "distance_threshold",
+                "features/per_frame/centroid_velocity_mag centroid_velocity_mag",
+                lambda x: np.nansum(x[x > 5], initial=0) / 30,
+            ),
+            FeatureInEvent(
+                "distance_seg",
+                "features/per_frame/shape_descriptor centroid_speed",
+                lambda x: np.nansum(x, initial=0) / 30,
+            ),
             FeatureInEvent("closest_id", "closest_identities", np.median),
             FeatureInEvent("closest_lixit", "closest_lixit", np.median),
             FeatureInEvent("closest_corner", "closest_corners", np.median),
@@ -1003,6 +1017,14 @@ def bouts_to_bins(
                 bins_to_summarize["calc_dist"] = (
                     bins_to_summarize["distance"] * bins_to_summarize["percent_bout"]
                 )
+                bins_to_summarize["calc_dist_threshold"] = (
+                    bins_to_summarize["distance_threshold"]
+                    * bins_to_summarize["percent_bout"]
+                )
+                bins_to_summarize["calc_dist_seg"] = (
+                    bins_to_summarize["distance_seg"]
+                    * bins_to_summarize["percent_bout"]
+                )
             else:
                 pass
             pd.options.mode.chained_assignment = "warn"
@@ -1070,6 +1092,12 @@ def bouts_to_bins(
                 results["behavior_dist"] = bins_to_summarize.loc[
                     bins_to_summarize["is_behavior"] == 1, "calc_dist"
                 ].sum()
+                results["behavior_dist_threshold"] = bins_to_summarize.loc[
+                    bins_to_summarize["is_behavior"] == 1, "calc_dist_threshold"
+                ].sum()
+                results["behavior_dist_seg"] = bins_to_summarize.loc[
+                    bins_to_summarize["is_behavior"] == 1, "calc_dist_seg"
+                ].sum()
             results_df_list.append(pd.DataFrame(results))
 
         # Remove an non-informative rows
@@ -1151,6 +1179,8 @@ def __init__(self, settings: ClassifierSettings, data: pd.DataFrame):
             "time",
             "not_behavior_dist",
             "behavior_dist",
+            "behavior_dist_threshold",
+            "behavior_dist_seg",
             "avg_bout_duration",
             "_stats_sample_count",
             "bout_duration_std",