Summary

app.py

@@ -53,7 +53,16 @@ with gr.Blocks(css="app.css") as demo:
                 output_statistics = gr.Plot(label="Statistics of emotions", elem_classes="stat")
         gr.Examples(
             ["videos/video1.mp4",
-            "videos/video2.mp4"],
+            "videos/video2.mp4",
+            "new_videos/01.mp4",
+            "new_videos/02.mp4",
+            "new_videos/14.mp4",
+            "new_videos/16.mp4",
+            "new_videos/20.mp4",
+            "new_videos/36.mp4",
+            "new_videos/38.mp4",
+            "new_videos/45.mp4",
+            ],
             [input_video],
         )
 

app/app_utils.py

@@ -44,11 +44,12 @@ def preprocess_image_and_predict(inp):
                 startX, startY, endX, endY = get_box(fl, w, h)
                 cur_face = inp[startY:endY, startX:endX]
                 cur_face_n = pth_processing(Image.fromarray(cur_face))
-                prediction = (
-                    torch.nn.functional.softmax(pth_model_static(cur_face_n), dim=1)
-                    .detach()
-                    .numpy()[0]
-                )
+                with torch.no_grad():
+                    prediction = (
+                        torch.nn.functional.softmax(pth_model_static(cur_face_n), dim=1)
+                        .detach()
+                        .numpy()[0]
+                    )
                 confidences = {DICT_EMO[i]: float(prediction[i]) for i in range(7)}
                 grayscale_cam = cam(input_tensor=cur_face_n)
                 grayscale_cam = grayscale_cam[0, :]
@@ -104,7 +105,8 @@ def preprocess_video_and_predict(video):
 
                     if count_face%config_data.FRAME_DOWNSAMPLING == 0:
                         cur_face_copy = pth_processing(Image.fromarray(cur_face))
-                        features = torch.nn.functional.relu(pth_model_static.extract_features(cur_face_copy)).detach().numpy()
+                        with torch.no_grad():
+                            features = torch.nn.functional.relu(pth_model_static.extract_features(cur_face_copy)).detach().numpy()
 
                         grayscale_cam = cam(input_tensor=cur_face_copy)
                         grayscale_cam = grayscale_cam[0, :]
@@ -120,7 +122,8 @@ def preprocess_video_and_predict(video):
 
                         lstm_f = torch.from_numpy(np.vstack(lstm_features))
                         lstm_f = torch.unsqueeze(lstm_f, 0)
-                        output = pth_model_dynamic(lstm_f).detach().numpy()
+                        with torch.no_grad():
+                            output = pth_model_dynamic(lstm_f).detach().numpy()
                         last_output = output
 
                         if count_face == 0: