Added tests for preprocess.py

deepof/preprocess.py

@@ -506,6 +506,7 @@ class table_dict(dict):
             {k: value for k, value in self.items() if k in keys}, self._type
         )
 
+    # noinspection PyTypeChecker
     def plot_heatmaps(self, bodyparts, save=False, i=0):
 
         if self._type != "coords" or self._polar:
@@ -549,6 +550,7 @@ class table_dict(dict):
 
         return X_train, X_test
 
+    # noinspection PyTypeChecker,PyGlobalUndefined
     def preprocess(
         self,
         window_size=1,
@@ -565,6 +567,7 @@ class table_dict(dict):
         """Builds a sliding window. If specified, splits train and test and
            Z-scores the data using sklearn's standard scaler"""
 
+        global g
         X_train, X_test = self.get_training_set(test_videos)
 
         if scale:
@@ -616,7 +619,7 @@ class table_dict(dict):
                 ]
             )
             g /= np.max(g)
-            X_train = X_train * g.reshape(1, window_size, 1)
+            X_train = X_train * g.reshape([1, window_size, 1])
 
         if test_videos:
 
@@ -629,7 +632,7 @@ class table_dict(dict):
                 X_test = align_trajectories(X_test, align)
 
             if conv_filter == "gaussian":
-                X_test = X_test * g.reshape(1, window_size, 1)
+                X_test = X_test * g.reshape([1, window_size, 1])
 
             if shuffle:
                 X_train = X_train[
@@ -650,7 +653,7 @@ class table_dict(dict):
 
     def random_projection(self, n_components=None, sample=1000):
 
-        X = self.get_training_set()
+        X = self.get_training_set()[0]
         X = X[np.random.choice(X.shape[0], sample, replace=False), :]
 
         rproj = random_projection.GaussianRandomProjection(n_components=n_components)
@@ -660,7 +663,7 @@ class table_dict(dict):
 
     def pca(self, n_components=None, sample=1000, kernel="linear"):
 
-        X = self.get_training_set()
+        X = self.get_training_set()[0]
         X = X[np.random.choice(X.shape[0], sample, replace=False), :]
 
         pca = KernelPCA(n_components=n_components, kernel=kernel)
@@ -670,7 +673,7 @@ class table_dict(dict):
 
     def tsne(self, n_components=None, sample=1000, perplexity=30):
 
-        X = self.get_training_set()
+        X = self.get_training_set()[0]
         X = X[np.random.choice(X.shape[0], sample, replace=False), :]
 
         tsne = TSNE(n_components=n_components, perplexity=perplexity)

deepof/utils.py

@@ -16,7 +16,7 @@ from scipy import spatial
 from scipy import stats
 from sklearn import mixture
 from tqdm import tqdm
-from typing import Tuple, Any, List, Union, Dict, NewType
+from typing import Tuple, Any, List, Union, NewType
 
 # DEFINE CUSTOM ANNOTATED TYPES #
 
@@ -936,7 +936,34 @@ def rule_based_tagging(
     huddle_spine: int = 10,
     huddle_speed: int = 1,
 ) -> pd.DataFrame:
-    """Outputs a dataframe with the motives registered per frame."""
+    """Outputs a dataframe with the registered motives per frame. If specified, produces a labeled
+    video displaying the information in real time
+
+    Parameters:
+        - tracks (list):
+        - videos (list):
+        - coordinates (deepof.preprocessing.coordinates):
+        - vid_index (int):
+        - animal_ids (list):
+        - show (bool):
+        - save (bool):
+        - fps (float):
+        - speed_pause (int):
+        - frame_limit (float):
+        - recog_limit (int):
+        - path (str):
+        - arena_type (str):
+        - close_contact_tol (int):
+        - side_contact_tol (int):
+        - follow_frames (int):
+        - follow_tol (int):
+        - huddle_forward (int):
+        - huddle_spine (int):
+        - huddle_speed (int):
+
+    Returns:
+        - tag_df (pandas.DataFrame): table with traits as columns and frames as rows. Each
+        value is a boolean indicating trait detection at a given time"""
 
     vid_name = re.findall("(.*?)_", tracks[vid_index])[0]
 

tests/test_preprocess.py

@@ -206,7 +206,7 @@ def test_get_table_dicts(nodes, ego, sampler):
     else:
         align = False
 
-    table.preprocess(
+    prep = table.preprocess(
         window_size=11,
         window_step=1,
         scale=sampler.draw(st.one_of(st.just("standard"), st.just("minmax"))),
@@ -218,3 +218,14 @@ def test_get_table_dicts(nodes, ego, sampler):
         shuffle=sampler.draw(st.booleans()),
         align=align,
     )
+
+    assert (type(prep) == np.ndarray or type(prep) == tuple)
+
+    if type(prep) == tuple:
+        assert type(prep[0]) == np.ndarray
+
+    # deepof dimensionality reduction testing
+
+    assert type(table.random_projection(n_components=2, sample=50)) == tuple
+    assert type(table.pca(n_components=2, sample=50)) == tuple
+    assert type(table.tsne(n_components=2, sample=50)) == tuple

tests/test_utils.py

@@ -260,6 +260,7 @@ def test_smooth_boolean_array(a):
     smooth = smooth_boolean_array(a)
 
     def trans(x):
+        """In situ function for computing boolean transitions"""
         return sum([i + 1 != i for i in range(x.shape[0] - 1)])
 
     assert trans(a) >= trans(smooth)