Update app.py

app.py

@@ -71,36 +71,52 @@ def make_prediction(image, taxonomic_decision, taxonomic_level):
     # Aggregate predictions based on the current taxonomic level
     aggregated_predictions, aggregated_class_labels = aggregate_predictions(prediction, taxonomic_levels[current_level_index], class_names)
 
-    # Check if the confidence of the top prediction meets the threshold
-    top_prediction_index = np.argmax(aggregated_predictions)
-    top_prediction_confidence = aggregated_predictions[0][top_prediction_index]
-
-    # If the user specified a taxonomic level, do not automatically promote to a higher level
+    # If the user specified a taxonomic level, simply get the highest prediction at that level
     if taxonomic_decision == "Yes, I want to specify the taxonomic level":
-        if current_level_index == 0 and top_prediction_confidence < 0.80:
-            return "<h1 style='font-weight: bold;'>Confidence too low for specified taxonomic level</h1>"
+        # Get the predicted class index for the current level
+        predicted_class_index = np.argmax(aggregated_predictions)
+        predicted_class_name = aggregated_class_labels[predicted_class_index]
+
+        # Construct the output message without considering confidence
+        output_text = f"<h1 style='font-weight: bold;'>{predicted_class_name}</h1>"
+        
+        # Add the top 5 predictions
+        output_text += "<h4 style='font-weight: bold; font-size: 1.2em;'>Top-5 predictions:</h4>"
+
+        top_indices = np.argsort(aggregated_predictions[0])[-5:][::-1]  # Get top 5 predictions
 
-    # Proceed to the next level if the confidence threshold is not met and the user allows it
+        for i in top_indices:
+            class_name = aggregated_class_labels[i]
+            confidence_percentage = aggregated_predictions[0][i] * 100
+            output_text += f"<div style='display: flex; justify-content: space-between;'>" \
+                           f"<span>{class_name}</span>" \
+                           f"<span style='margin-left: auto;'>{confidence_percentage:.2f}%</span></div>"
+
+        return output_text
+
+    # Confidence checking for the automatic model decision
+    # Loop through taxonomic levels if the user lets the model decide
     while current_level_index < len(taxonomic_levels):
-        # Check if confidence is above the threshold at the current level
+        # Aggregate predictions for the next level
+        aggregated_predictions, aggregated_class_labels = aggregate_predictions(prediction, taxonomic_levels[current_level_index], class_names)
+        
+        # Check if the confidence of the top prediction meets the threshold
+        top_prediction_index = np.argmax(aggregated_predictions)
+        top_prediction_confidence = aggregated_predictions[0][top_prediction_index]
+
         if top_prediction_confidence >= 0.80:
             break  # Confidence threshold met, exit loop
 
         current_level_index += 1  # Move to the next taxonomic level
-        if current_level_index < len(taxonomic_levels):
-            # Aggregate predictions for the next level
-            aggregated_predictions, aggregated_class_labels = aggregate_predictions(prediction, taxonomic_levels[current_level_index], class_names)
-            top_prediction_index = np.argmax(aggregated_predictions)
-            top_prediction_confidence = aggregated_predictions[0][top_prediction_index]
 
     # Check if a valid prediction was made
     if current_level_index == len(taxonomic_levels):
         return "<h1 style='font-weight: bold;'>Unknown animal</h1>"  # No valid predictions met the confidence criteria
-    
+
     # Get the predicted class name for the top prediction
     predicted_class_index = np.argmax(aggregated_predictions)
     predicted_class_name = aggregated_class_labels[predicted_class_index]
-    
+
     # Check if common name should be displayed (only at species level)
     if taxonomic_levels[current_level_index] == "species":
         predicted_common_name = taxo_df[taxo_df[taxonomic_levels[current_level_index]] == predicted_class_name]['common_name'].values[0]