Inverse predictions

.gitattributes

@@ -1,3 +1,4 @@
 *.ipynb filter=nbstripout
 *.zpln filter=nbstripout
 *.ipynb diff=ipynb
+* text=auto

docs/notebooks/gait/gait_analysis.ipynb

@@ -15,7 +15,7 @@
    "source": [
     "import os\n",
     "from paradigma.preprocessing_config import IMUPreprocessingConfig\n",
-    "from paradigma.gait.gait_analysis import extract_gait_features_io, detect_gait_io, extract_arm_activity_features_io, detect_other_arm_activities_io, quantify_arm_swing_io\n",
+    "from paradigma.gait.gait_analysis import extract_gait_features_io, detect_gait_io, extract_arm_activity_features_io, filter_gait_io, quantify_arm_swing_io\n",
     "from paradigma.gait.gait_analysis_config import GaitFeatureExtractionConfig, GaitDetectionConfig, ArmActivityFeatureExtractionConfig, FilteringGaitConfig, ArmSwingQuantificationConfig\n",
     "from paradigma.imu_preprocessing import preprocess_imu_data_io"
    ]
@@ -126,7 +126,7 @@
    "outputs": [],
    "source": [
     "config = FilteringGaitConfig()\n",
-    "detect_other_arm_activities_io(path_to_extracted_features, path_to_predictions, path_to_classifier, config)"
+    "filter_gait_io(path_to_extracted_features, path_to_predictions, path_to_classifier, config)"
    ]
   },
   {

src/paradigma/constants.py

@@ -21,8 +21,8 @@ class DataColumns():
     GRAV_ACCELEROMETER_Z : str = "grav_accelerometer_z"
     PRED_GAIT_PROBA: str = "pred_gait_proba"
     PRED_GAIT : str = "pred_gait"
-    PRED_OTHER_ARM_ACTIVITY_PROBA: str = "pred_other_arm_activity_proba"
-    PRED_OTHER_ARM_ACTIVITY : str = "pred_other_arm_activity"
+    PRED_NO_OTHER_ARM_ACTIVITY_PROBA: str = "pred_no_other_arm_activity_proba"
+    PRED_NO_OTHER_ARM_ACTIVITY : str = "pred_no_other_arm_activity"
     ANGLE : str = "angle"
     ANGLE_SMOOTH : str = "angle_smooth"
     VELOCITY : str = "velocity"

src/paradigma/gait/gait_analysis.py

@@ -287,7 +287,7 @@ def extract_arm_activity_features_io(input_path: Union[str, Path], output_path:
     write_df_data(metadata_time, metadata_samples, output_path, 'arm_activity_meta.json', df_windowed)
 
 
-def detect_other_arm_activities(df: pd.DataFrame, config: FilteringGaitConfig, clf: Union[LogisticRegression, RandomForestClassifier]) -> pd.DataFrame:
+def filter_gait(df: pd.DataFrame, config: FilteringGaitConfig, clf: Union[LogisticRegression, RandomForestClassifier]) -> pd.DataFrame:
 
     # Prepare the data
     clf.feature_names_in_ = ['std_norm_acc'] + [f'{x}_power_below_gait' for x in config.l_accelerometer_cols] + \
@@ -302,25 +302,25 @@ def detect_other_arm_activities(df: pd.DataFrame, config: FilteringGaitConfig, c
     X = df.loc[:, clf.feature_names_in_]
 
     # Make prediction
-    df[DataColumns.PRED_OTHER_ARM_ACTIVITY_PROBA] = clf.predict_proba(X)[:, 1]
+    df[DataColumns.PRED_NO_OTHER_ARM_ACTIVITY_PROBA] = clf.predict_proba(X)[:, 1]
 
     return df
 
-def detect_other_arm_activities_io(input_path: Union[str, Path], output_path: Union[str, Path], path_to_classifier_input: Union[str, Path], config: FilteringGaitConfig) -> None:
+def filter_gait_io(input_path: Union[str, Path], output_path: Union[str, Path], path_to_classifier_input: Union[str, Path], config: FilteringGaitConfig) -> None:
     # Load the data
     metadata_time, metadata_samples = read_metadata(input_path, config.meta_filename, config.time_filename, config.values_filename)
     df = tsdf.load_dataframe_from_binaries([metadata_time, metadata_samples], tsdf.constants.ConcatenationType.columns)
 
     # Load the classifier
     clf = pd.read_pickle(os.path.join(path_to_classifier_input, 'classifiers', config.classifier_file_name))
 
-    df = detect_other_arm_activities(df, config, clf)
+    df = filter_gait(df, config, clf)
 
     # Prepare the metadata
     metadata_samples.file_name = 'arm_activity_values.bin'
     metadata_time.file_name = 'arm_activity_time.bin'
 
-    metadata_samples.channels = [DataColumns.PRED_OTHER_ARM_ACTIVITY_PROBA]
+    metadata_samples.channels = [DataColumns.PRED_NO_OTHER_ARM_ACTIVITY_PROBA]
     metadata_samples.units = ['probability']
 
     metadata_time.channels = [DataColumns.TIME]
@@ -332,11 +332,11 @@ def detect_other_arm_activities_io(input_path: Union[str, Path], output_path: Un
 def quantify_arm_swing(df: pd.DataFrame, config: ArmSwingQuantificationConfig) -> pd.DataFrame:
 
     # temporarily for testing: manually determine predictions
-    df[DataColumns.PRED_OTHER_ARM_ACTIVITY_PROBA] = np.concatenate([np.repeat([1], df.shape[0]//3), np.repeat([0], df.shape[0]//3), np.repeat([1], df.shape[0] - 2*df.shape[0]//3)], axis=0)
+    df[DataColumns.PRED_NO_OTHER_ARM_ACTIVITY_PROBA] = np.concatenate([np.repeat([1], df.shape[0]//3), np.repeat([0], df.shape[0]//3), np.repeat([1], df.shape[0] - 2*df.shape[0]//3)], axis=0)
 
     # keep only predicted arm swing
     # TODO: Aggregate overlapping windows for probabilities
-    df_filtered = df.loc[df[DataColumns.PRED_OTHER_ARM_ACTIVITY_PROBA]>=0.5].copy().reset_index(drop=True)
+    df_filtered = df.loc[df[DataColumns.PRED_NO_OTHER_ARM_ACTIVITY_PROBA]>=0.5].copy().reset_index(drop=True)
 
     del df
 
@@ -398,7 +398,7 @@ def quantify_arm_swing_io(path_to_feature_input: Union[str, Path], path_to_predi
     df_features = df_features[l_feature_cols]
 
     # Concatenate features and predictions
-    df = pd.concat([df_features, df_predictions[DataColumns.PRED_OTHER_ARM_ACTIVITY_PROBA]], axis=1)
+    df = pd.concat([df_features, df_predictions[DataColumns.PRED_NO_OTHER_ARM_ACTIVITY_PROBA]], axis=1)
 
     df_aggregates = quantify_arm_swing(df, config)
 

src/paradigma/gait/gait_analysis_config.py

@@ -267,8 +267,8 @@ def __init__(self) -> None:
 
         self.angle_colname = DataColumns.ANGLE
         self.velocity_colname = DataColumns.VELOCITY
-        self.pred_other_arm_activity_proba_colname = DataColumns.PRED_OTHER_ARM_ACTIVITY_PROBA
-        self.pred_other_arm_activity_colname = DataColumns.PRED_OTHER_ARM_ACTIVITY
+        self.pred_other_arm_activity_proba_colname = DataColumns.PRED_NO_OTHER_ARM_ACTIVITY_PROBA
+        self.pred_other_arm_activity_colname = DataColumns.PRED_NO_OTHER_ARM_ACTIVITY
 
         self.window_length_s = 3
         self.window_step_size = 0.75

tests/data/4.predictions/gait/arm_activity_meta.json

@@ -26,7 +26,7 @@
         {
             "file_name": "arm_activity_values.bin",
             "channels": [
-                "pred_other_arm_activity_proba"
+                "pred_no_other_arm_activity_proba"
             ],
             "units": [
                 "probability"

tests/test_gait_analysis.py

@@ -1,6 +1,6 @@
 from pathlib import Path
 
-from paradigma.gait.gait_analysis import detect_other_arm_activities_io, detect_gait_io, extract_arm_activity_features_io, extract_gait_features_io, quantify_arm_swing_io
+from paradigma.gait.gait_analysis import filter_gait_io, detect_gait_io, extract_arm_activity_features_io, extract_gait_features_io, quantify_arm_swing_io
 from paradigma.gait.gait_analysis_config import FilteringGaitConfig, ArmActivityFeatureExtractionConfig, ArmSwingQuantificationConfig, GaitDetectionConfig, GaitFeatureExtractionConfig
 from paradigma.imu_preprocessing import preprocess_imu_data_io
 from paradigma.preprocessing_config import IMUPreprocessingConfig
@@ -109,7 +109,7 @@ def test_5_arm_swing_detection_output(shared_datadir: Path):
     tested_output_path = reference_output_path / "test-output"
 
     config = FilteringGaitConfig()
-    detect_other_arm_activities_io(input_path, tested_output_path, path_to_classifier_input, config)
+    filter_gait_io(input_path, tested_output_path, path_to_classifier_input, config)
     compare_data(reference_output_path, tested_output_path, arm_activity_binaries_pairs)