1

app.py

@@ -3,6 +3,10 @@ import gradio as gr
 from tabs.FACS_analysis import create_facs_analysis_tab
 from ui_components import CUSTOM_CSS, HEADER_HTML, DISCLAIMER_HTML
 
+# Move load_example to a global scope
+def load_example():
+    pass  # Your logic here
+
 # Define the tab structure
 TAB_STRUCTURE = [
     ("Visual Analysis", [

app/model.py

@@ -49,18 +49,21 @@ def load_models():
     # Load the static ResNet50 model
     pth_model_static = load_model(ResNet50, STATIC_MODEL_PATH)
     
-    # Define LSTMPyTorch parameters (set the correct values for your model)
+    # Define LSTMPyTorch parameters
     input_size = 2048  # Example value: This should match the feature size from the ResNet50 output
     hidden_size = 512  # Example value: Adjust based on your LSTM architecture
     num_layers = 2     # Example value: Number of layers in the LSTM
     num_classes = 7    # Example value: Number of emotion classes
 
-    # Load the dynamic LSTM model with the correct arguments
-    pth_model_dynamic = load_model(LSTMPyTorch, DYNAMIC_MODEL_PATH, input_size, hidden_size, num_layers, num_classes)
+    # Load the dynamic LSTM model (if available)
+    pth_model_dynamic = None
+    if os.path.exists(DYNAMIC_MODEL_PATH):
+        pth_model_dynamic = load_model(LSTMPyTorch, DYNAMIC_MODEL_PATH, input_size, hidden_size, num_layers, num_classes)
+    else:
+        logger.error(f"Dynamic model file not found: {DYNAMIC_MODEL_PATH}")
     
     # Initialize GradCAM
     cam = GradCAM(model=pth_model_static, target_layers=[pth_model_static.resnet.layer4])
     
     return pth_model_static, pth_model_dynamic, cam
 
-# Optionally, additional utility functions for model processing can be added here.