update app v3

changes due to uncompatible between deepface and tensorflow

app.py

@@ -5,12 +5,25 @@ import pandas as pd
 from datetime import datetime
 import time
 from transformers import pipeline
-from deepface import DeepFace
 import librosa
 from python_speech_features import mfcc
+import onnxruntime as ort
+import requests
+import os
+
+# Download emotion recognition ONNX model
+MODEL_URL = "https://github.com/onnx/models/raw/main/vision/body_analysis/emotion_ferplus/model/emotion-ferplus-8.onnx"
+MODEL_PATH = "emotion-ferplus-8.onnx"
+
+if not os.path.exists(MODEL_PATH):
+    response = requests.get(MODEL_URL)
+    with open(MODEL_PATH, "wb") as f:
+        f.write(response.content)
 
 # Initialize models
 voice_classifier = pipeline("audio-classification", model="superb/hubert-base-superb-er")
+emotion_session = ort.InferenceSession(MODEL_PATH)
+emotion_labels = ['neutral', 'happy', 'surprise', 'sad', 'angry', 'disgust', 'fear', 'contempt']
 
 # Global variables to store results
 emotion_history = []
@@ -18,22 +31,38 @@ current_emotions = {"face": "neutral", "voice": "neutral"}
 last_update_time = time.time()
 
 def analyze_face(frame):
-    """Analyze facial expressions in the frame using DeepFace"""
+    """Analyze facial expressions in the frame using ONNX model"""
     try:
-        # Convert frame to RGB (DeepFace expects RGB)
-        rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+        # Preprocess frame
+        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+        face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
+        faces = face_cascade.detectMultiScale(gray, 1.3, 5)
         
-        # Detect emotions
-        result = DeepFace.analyze(rgb_frame, actions=['emotion'], enforce_detection=False)
-        
-        if result and isinstance(result, list):
-            emotions = result[0]['emotion']
-            dominant_emotion = max(emotions, key=emotions.get)
+        if len(faces) > 0:
+            x, y, w, h = faces[0]
+            face_roi = gray[y:y+h, x:x+w]
+            face_roi = cv2.resize(face_roi, (64, 64))
+            face_roi = face_roi.astype('float32') / 255.0
+            face_roi = np.expand_dims(face_roi, axis=0)
+            face_roi = np.expand_dims(face_roi, axis=0)
+            
+            # Run inference
+            input_name = emotion_session.get_inputs()[0].name
+            output_name = emotion_session.get_outputs()[0].name
+            results = emotion_session.run([output_name], {input_name: face_roi})[0]
+            
+            # Get emotion probabilities
+            emotion_probs = results[0]
+            dominant_emotion = emotion_labels[np.argmax(emotion_probs)]
+            
+            # Create emotion dictionary
+            emotions = {label: float(prob) for label, prob in zip(emotion_labels, emotion_probs)}
             return dominant_emotion, emotions
-        return "neutral", {"angry": 0, "disgust": 0, "fear": 0, "happy": 0, "sad": 0, "surprise": 0, "neutral": 1}
+        
+        return "neutral", {label: 0.0 for label in emotion_labels}
     except Exception as e:
         print(f"Face analysis error: {e}")
-        return "neutral", {"angry": 0, "disgust": 0, "fear": 0, "happy": 0, "sad": 0, "surprise": 0, "neutral": 1}
+        return "neutral", {label: 0.0 for label in emotion_labels}
 
 def analyze_voice(audio):
     """Analyze voice tone from audio"""