Update app.py

app.py

@@ -1,19 +1,17 @@
 import streamlit as st
 import tensorflow as tf
-from tensorflow import keras
 import numpy as np
 from PIL import Image
-import io
 import cv2
 import openai
 import os
+import io
 
 # Set up OpenAI API key
 openai.api_key = "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"
 
-# Load pre-trained models
+# Load pre-trained model for face detection
 face_detection_model = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
-emotion_model = keras.models.load_model('path_to_emotion_model.h5')  # You'll need to train or download this model
 
 def detect_face(image):
     gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
@@ -31,20 +29,26 @@ def analyze_symmetry(face):
     symmetry_score = 1 - (np.sum(diff) / (255 * height * (width//2)))
     return symmetry_score
 
-def analyze_emotion(face):
-    face = cv2.resize(face, (48, 48))
-    face = cv2.cvtColor(face, cv2.COLOR_BGR2GRAY)
-    face = np.reshape(face, [1, 48, 48, 1]) / 255.0
-    emotion_pred = emotion_model.predict(face)
-    emotions = ['Angry', 'Disgust', 'Fear', 'Happy', 'Sad', 'Surprise', 'Neutral']
-    return emotions[np.argmax(emotion_pred)]
-
 def analyze_aesthetic_quality(image):
+    # Convert the image to a byte stream
+    img_byte_arr = io.BytesIO()
+    Image.fromarray(image).save(img_byte_arr, format='PNG')
+    img_byte_arr = img_byte_arr.getvalue()
+
+    # Encode the image
+    encoded_image = base64.b64encode(img_byte_arr).decode('ascii')
+
     prompt = f"Analyze the aesthetic quality of this image, focusing on lighting, composition, and overall visual appeal. Provide a brief description and a score out of 10."
     response = openai.ChatCompletion.create(
         model="gpt-4-vision-preview",
         messages=[
-            {"role": "user", "content": [{"type": "text", "text": prompt}, {"type": "image_url", "image_url": {"url": image}}]}
+            {
+                "role": "user", 
+                "content": [
+                    {"type": "text", "text": prompt},
+                    {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{encoded_image}"}}
+                ]
+            }
         ]
     )
     return response.choices[0].message.content
@@ -59,21 +63,16 @@ def compare_photos(image1, image2):
     symmetry1 = analyze_symmetry(face1)
     symmetry2 = analyze_symmetry(face2)
     
-    emotion1 = analyze_emotion(face1)
-    emotion2 = analyze_emotion(face2)
-    
     aesthetic1 = analyze_aesthetic_quality(image1)
     aesthetic2 = analyze_aesthetic_quality(image2)
     
     return {
         "Image 1": {
             "Symmetry": symmetry1,
-            "Emotion": emotion1,
             "Aesthetic Quality": aesthetic1
         },
         "Image 2": {
             "Symmetry": symmetry2,
-            "Emotion": emotion2,
             "Aesthetic Quality": aesthetic2
         }
     }
@@ -104,12 +103,10 @@ def main():
                 with col1:
                     st.write("Image 1")
                     st.write(f"Symmetry: {results['Image 1']['Symmetry']:.2f}")
-                    st.write(f"Emotion: {results['Image 1']['Emotion']}")
                     st.write(f"Aesthetic Quality: {results['Image 1']['Aesthetic Quality']}")
                 with col2:
                     st.write("Image 2")
                     st.write(f"Symmetry: {results['Image 2']['Symmetry']:.2f}")
-                    st.write(f"Emotion: {results['Image 2']['Emotion']}")
                     st.write(f"Aesthetic Quality: {results['Image 2']['Aesthetic Quality']}")
                 
                 st.write("Note: Beauty is subjective, and this tool is for entertainment purposes only.")