Update app.py

app.py

@@ -29,9 +29,9 @@ matplotlib.rcParams['savefig.dpi'] = 400
 # Initialize models and other global variables
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
-FIXED_FPS = 20
+FIXED_FPS = 5
 
-mtcnn = MTCNN(keep_all=False, device=device, thresholds=[0.95, 0.95, 0.95], min_face_size=100)
+mtcnn = MTCNN(keep_all=False, device=device, thresholds=[0.95, 0.95, 0.95], min_face_size=80)
 model = InceptionResnetV1(pretrained='vggface2').eval().to(device)
 
 mp_face_mesh = mp.solutions.face_mesh
@@ -215,7 +215,10 @@ def process_frames(frames_folder, aligned_faces_folder, frame_count, progress, b
                     if face.size > 0:
                         results = face_mesh.process(cv2.cvtColor(face, cv2.COLOR_BGR2RGB))
                         if results.multi_face_landmarks and is_frontal_face(results.multi_face_landmarks[0].landmark):
-                            aligned_face = alignFace(face)
+
+                            #aligned_face = alignFace(face)
+                            aligned_face = face
+
                             if aligned_face is not None:
                                 aligned_face_resized = cv2.resize(aligned_face, (160, 160))
                                 output_path = os.path.join(aligned_faces_folder, f"frame_{frame_num}_face.jpg")
@@ -290,22 +293,22 @@ class Autoencoder(nn.Module):
     def __init__(self, input_size):
         super(Autoencoder, self).__init__()
         self.encoder = nn.Sequential(
-            nn.Linear(input_size, 512),
-            nn.ReLU(),
-            nn.Linear(512, 256),
+            nn.Linear(input_size, 256),
             nn.ReLU(),
             nn.Linear(256, 128),
             nn.ReLU(),
-            nn.Linear(128, 64)
+            nn.Linear(128, 64),
+            nn.ReLU(),
+            nn.Linear(64, 32)
         )
         self.decoder = nn.Sequential(
+            nn.Linear(32, 64),
+            nn.ReLU(),
             nn.Linear(64, 128),
             nn.ReLU(),
             nn.Linear(128, 256),
             nn.ReLU(),
-            nn.Linear(256, 512),
-            nn.ReLU(),
-            nn.Linear(512, input_size)
+            nn.Linear(256, input_size)
         )
 
     def forward(self, x):
@@ -321,7 +324,7 @@ def determine_anomalies(mse_values, threshold):
     anomalies = mse_values > (mean + threshold * std)
     return anomalies
 
-def anomaly_detection(X_embeddings, X_posture, epochs=200, batch_size=8, patience=3):
+def anomaly_detection(X_embeddings, X_posture, epochs=200, batch_size=8, patience=5):
     # Normalize posture
     scaler_posture = MinMaxScaler()
     X_posture_scaled = scaler_posture.fit_transform(X_posture.reshape(-1, 1))
@@ -519,8 +522,8 @@ def plot_posture(df, posture_scores, color='blue', anomaly_threshold=4):
 
 
 def plot_mse_heatmap(mse_values, title, df):
-    plt.figure(figsize=(20, 10), dpi=400)
-    fig, ax = plt.subplots(figsize=(20, 10))
+    plt.figure(figsize=(20, 5), dpi=400)
+    fig, ax = plt.subplots(figsize=(20, 5))
 
     # Reshape MSE values to 2D array for heatmap
     mse_2d = mse_values.reshape(1, -1)
@@ -626,7 +629,7 @@ def draw_pose_landmarks(frame, landmarks):
 
     return annotated_frame
 
-def get_all_face_samples(organized_faces_folder, output_folder, largest_cluster, max_samples=400):
+def get_all_face_samples(organized_faces_folder, output_folder, largest_cluster, max_samples=500):
     face_samples = {"most_frequent": [], "others": []}
     for cluster_folder in sorted(os.listdir(organized_faces_folder)):
         if cluster_folder.startswith("person_"):
@@ -836,13 +839,13 @@ with gr.Blocks() as iface:
     with gr.Row():
         video_input = gr.Video()
 
-    anomaly_threshold = gr.Slider(minimum=1, maximum=5, step=0.1, value=3.5, label="Anomaly Detection Threshold")
+    anomaly_threshold = gr.Slider(minimum=1, maximum=5, step=0.1, value=3, label="Anomaly Detection Threshold")
     process_btn = gr.Button("Process Video")
     progress_bar = gr.Progress()
     execution_time = gr.Number(label="Execution Time (seconds)")
 
     with gr.Group(visible=False) as results_group:
-        results_text = gr.TextArea(label="Anomaly Detection Results", lines=6)
+        results_text = gr.TextArea(label="Anomaly Detection Results", lines=4)
 
         with gr.Tab("Facial Features"):
             mse_features_plot = gr.Plot(label="MSE: Facial Features")