Merge pull request #86 from cosanlab/hog_aus

Hog aus

Author: jcheong0428

feat/au_detectors/DRML/DRML_model.py

@@ -91,7 +91,6 @@ def forward(self, x):
 
 if __name__ == "__main__":
     from torch.autograd import Variable
-
     pic_x = Variable(torch.rand(8, 3, 170, 170))
     net = DRML_net(AU_num=12)
     print(net)

feat/au_detectors/DRML/DRML_test.py

@@ -58,7 +58,6 @@ def detect_au(self, imgs, landmarks=None):
         pred_au = self.drml_net(input)
         all_pred_au = pred_au.data.cpu().float()
         # all_pred_au = (all_pred_au[:, 1, :]).exp()
-
         # all_pred_au[all_pred_au < 0.5] = 0
         # all_pred_au[all_pred_au >= 0.5] = 1
         all_pred_au = all_pred_au.data.numpy()

feat/au_detectors/StatLearning/SL_test.py

@@ -5,96 +5,111 @@
 import numpy as np
 from sklearn.preprocessing import StandardScaler
 from sklearn.decomposition import PCA
-from sklearn.metrics import classification_report,f1_score
+from sklearn.metrics import classification_report, f1_score
 from sklearn.svm import LinearSVC, SVC
 from sklearn.model_selection import GridSearchCV, RandomizedSearchCV
 from sklearn.ensemble import RandomForestClassifier
 from sklearn.datasets import make_classification
 from sklearn.linear_model import LogisticRegression
 from feat.utils import get_resource_path
 import joblib
-import os 
+import os
 
 
 def load_classifier(cf_path):
     clf = joblib.load(cf_path)
-    return clf 
+    return clf
+
 
 class RandomForestClassifier():
     def __init__(self) -> None:
-        self.pca_model = load_classifier(os.path.join(get_resource_path(),"hog_pca_all_emotio.joblib"))
-        self.classifier = load_classifier(os.path.join(get_resource_path(), "RF_568.joblib"))
-        self.scaler = load_classifier(os.path.join(get_resource_path(), "hog_scalar_aus.joblib"))
+        self.pca_model = load_classifier(os.path.join(
+            get_resource_path(), "hog_pca_all_emotio.joblib"))
+        self.classifier = load_classifier(
+            os.path.join(get_resource_path(), "RF_568.joblib"))
+        self.scaler = load_classifier(os.path.join(
+            get_resource_path(), "hog_scalar_aus.joblib"))
 
     def detect_au(self, frame, landmarks):
         """
         Note that here frame is represented by hogs
         """
         if len(frame.shape) < 2:
-            frame = frame.reshape(1,-1)
+            frame = frame.reshape(1, -1)
         if len(landmarks.shape) > 1:
-            landmarks = landmarks.flatten().reshape(1,-1)
-        
-        pca_transformed_frame = self.pca_model.transform(self.scaler.fit_transform(frame))
-        feature_cbd = np.concatenate((pca_transformed_frame,landmarks),1)
+            landmarks = landmarks.flatten().reshape(1, -1)
+
+        pca_transformed_frame = self.pca_model.transform(
+            self.scaler.fit_transform(frame))
+        feature_cbd = np.concatenate((pca_transformed_frame, landmarks), 1)
         pred_aus = []
         for keys in self.classifier:
             au_pred = self.classifier[keys].predict_proba(feature_cbd)
-            au_pred = au_pred[0,1]
+            au_pred = au_pred[0, 1]
             pred_aus.append(au_pred)
 
-        pred_aus = np.array(pred_aus).reshape(1,-1)
+        pred_aus = np.array(pred_aus).reshape(1, -1)
         return pred_aus
 
 
 class SVMClassifier():
     def __init__(self) -> None:
-        self.pca_model = load_classifier(os.path.join(get_resource_path(),"hog_pca_all_emotio.joblib"))
-        self.classifier = load_classifier(os.path.join(get_resource_path(),"svm_568.joblib"))
-        self.scaler = load_classifier(os.path.join(get_resource_path(), "hog_scalar_aus.joblib"))
+        self.pca_model = load_classifier(os.path.join(
+            get_resource_path(), "hog_pca_all_emotio.joblib"))
+        self.classifier = load_classifier(
+            os.path.join(get_resource_path(), "svm_568.joblib"))
+        self.scaler = load_classifier(os.path.join(
+            get_resource_path(), "hog_scalar_aus.joblib"))
 
     def detect_au(self, frame, landmarks):
         """
         Note that here frame is represented by hogs
         """
         if len(frame.shape) < 2:
-            frame = frame.reshape(1,-1)
+            frame = frame.reshape(1, -1)
         if len(landmarks.shape) > 1:
-            landmarks = landmarks.flatten().reshape(1,-1)
-        
-        pca_transformed_frame = self.pca_model.transform(self.scaler.fit_transform(frame))
-        feature_cbd = np.concatenate((pca_transformed_frame,landmarks),1)
+            landmarks = landmarks.flatten().reshape(1, -1)
+
+        pca_transformed_frame = self.pca_model.transform(
+            self.scaler.fit_transform(frame))
+        feature_cbd = np.concatenate((pca_transformed_frame, landmarks), 1)
         pred_aus = []
         for keys in self.classifier:
             au_pred = self.classifier[keys].predict(feature_cbd)
-            au_pred = au_pred[0] # probably need to delete this
+            au_pred = au_pred[0]  # probably need to delete this
             pred_aus.append(au_pred)
 
-        pred_aus = np.array(pred_aus).reshape(1,-1)
+        pred_aus = np.array(pred_aus).reshape(1, -1)
         return pred_aus
 
+
 class LogisticClassifier():
-    
+
     def __init__(self) -> None:
-        self.pca_model = load_classifier(os.path.join(get_resource_path(),"hog_pca_all_emotio.joblib"))
-        self.classifier = load_classifier(os.path.join(get_resource_path(),"Logistic_520.joblib"))
-        self.scaler = load_classifier(os.path.join(get_resource_path(), "hog_scalar_aus.joblib"))
+        self.pca_model = load_classifier(os.path.join(
+            get_resource_path(), "hog_pca_all_emotio.joblib"))
+        self.classifier = load_classifier(os.path.join(
+            get_resource_path(), "Logistic_520.joblib"))
+        self.scaler = load_classifier(os.path.join(
+            get_resource_path(), "hog_scalar_aus.joblib"))
+
     def detect_au(self, frame, landmarks):
         """
         Note that here frame is represented by hogs
         """
         if len(frame.shape) < 2:
-            frame = frame.reshape(1,-1)
+            frame = frame.reshape(1, -1)
         if len(landmarks.shape) > 1:
-            landmarks = landmarks.flatten().reshape(1,-1)
-        
-        pca_transformed_frame = self.pca_model.transform(self.scaler.fit_transform(frame))
-        feature_cbd = np.concatenate((pca_transformed_frame,landmarks),1)
+            landmarks = landmarks.flatten().reshape(1, -1)
+
+        pca_transformed_frame = self.pca_model.transform(
+            self.scaler.fit_transform(frame))
+        feature_cbd = np.concatenate((pca_transformed_frame, landmarks), 1)
         pred_aus = []
         for keys in self.classifier:
             au_pred = self.classifier[keys].predict_proba(feature_cbd)
-            au_pred = au_pred[0,1]
+            au_pred = au_pred[0, 1]
             pred_aus.append(au_pred)
 
-        pred_aus = np.array(pred_aus).reshape(1,-1)
+        pred_aus = np.array(pred_aus).reshape(1, -1)
         return pred_aus