Update app.py

app.py

@@ -7,6 +7,8 @@ import streamlit as st
 from PIL import Image
 import numpy as np
 import time
+from streamlit_webrtc import webrtc_streamer, VideoProcessorBase, RTCConfiguration
+import av
 
 # Définition du modèle CNN
 class EmotionCNN(nn.Module):
@@ -127,7 +129,7 @@ st.title("🎭 Détecteur d'Émotions en Temps Réel")
 def load_model():
     device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
     model = EmotionCNN().to(device)
-    model.load_state_dict(torch.load("cnn_emotion_model.pth", map_location=device))
+    model.load_state_dict(torch.load("modele/cnn_emotion_model.pth", map_location=device))
     model.eval()
     return model, device
 
@@ -150,85 +152,78 @@ def detect_faces(frame):
     faces = face_cascade.detectMultiScale(gray, 1.1, 4)
     return faces
 
-# Configuration de la webcam
+# Configuration RTC pour WebRTC (utile pour Hugging Face Spaces)
+RTC_CONFIGURATION = RTCConfiguration({
+    "iceServers": [{"urls": ["stun:stun.l.google.com:19302"]}]
+})
+
+# Classe pour traiter les frames vidéo
+class VideoProcessor(VideoProcessorBase):
+    def __init__(self):
+        self.model = model
+        self.device = device
+        self.transform = transform
+        self.face_cascade = face_cascade
+        self.emotion_dict = emotion_dict
+        self.emotion_placeholder = st.session_state.get('emotion_placeholder')
+        self.message_placeholder = st.session_state.get('message_placeholder')
+
+    def recv(self, frame):
+        img = frame.to_ndarray(format="bgr24")
+        faces = detect_faces(img)
+        
+        for (x, y, w, h) in faces:
+            cv2.rectangle(img, (x, y), (x+w, y+h), (0, 255, 0), 2)
+            face_img = img[y:y+h, x:x+w]
+            pil_img = Image.fromarray(cv2.cvtColor(face_img, cv2.COLOR_BGR2RGB))
+            img_tensor = self.transform(pil_img).unsqueeze(0).to(self.device)
+            with torch.no_grad():
+                output = self.model(img_tensor)
+                _, predicted = torch.max(output, 1)
+                emotion_idx = predicted.item()
+            emotion_name = self.emotion_dict[emotion_idx]["name"]
+            cv2.putText(img, emotion_name, (x, y-10),
+                        cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2)
+            
+            # Mettre à jour les placeholders (utiliser st.session_state pour partager)
+            if self.emotion_placeholder:
+                self.emotion_placeholder.markdown(f"""
+                <div class="emotion-box">
+                    <div class="emotion-title">{emotion_name}</div>
+                </div>
+                """, unsafe_allow_html=True)
+            
+            if self.message_placeholder:
+                self.message_placeholder.markdown(f"""
+                <div class="emotion-box">
+                    <div class="emotion-message">{self.emotion_dict[emotion_idx]["message"]}</div>
+                </div>
+                """, unsafe_allow_html=True)
+
+        return av.VideoFrame.from_ndarray(img, format="bgr24")
+
+# Configuration de la webcam avec streamlit-webrtc
 col1, col2 = st.columns([2, 1])
 
 with col1:
     st.markdown("### 📹 Flux Vidéo")
-    frame_placeholder = st.empty()
+    # Lancer le flux webcam
+    webrtc_ctx = webrtc_streamer(
+        key="emotion-detection",
+        rtc_configuration=RTC_CONFIGURATION,
+        video_processor_factory=VideoProcessor,
+        media_stream_constraints={"video": True, "audio": False},
+    )
 
 with col2:
     st.markdown("### 😊 Émotion Détectée")
-    emotion_placeholder = st.empty()
-    message_placeholder = st.empty()
-
-# Bouton pour démarrer/arrêter la détection
-start_button = st.button("Démarrer la Détection")
-
-if start_button:
-    cap = cv2.VideoCapture(0)
-    
-    if not cap.isOpened():
-        st.error("❌ Impossible d'accéder à la webcam. Veuillez vérifier vos permissions.")
-        st.stop()
-    
-    try:
-        while True:
-            ret, frame = cap.read()
-            if not ret:
-                st.error("❌ Erreur lors de la capture vidéo.")
-                break
-
-            # Détection des visages
-            faces = detect_faces(frame)
-            
-            # Traitement de chaque visage détecté
-            for (x, y, w, h) in faces:
-                # Dessiner la bounding box
-                cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 2)
-                
-                # Extraire et prétraiter le visage
-                face_img = frame[y:y+h, x:x+w]
-                pil_img = Image.fromarray(cv2.cvtColor(face_img, cv2.COLOR_BGR2RGB))
-                
-                # Prédiction de l'émotion
-                img_tensor = transform(pil_img).unsqueeze(0).to(device)
-                with torch.no_grad():
-                    output = model(img_tensor)
-                    _, predicted = torch.max(output, 1)
-                    emotion_idx = predicted.item()
-                    
-                # Afficher l'émotion sur l'image
-                emotion_name = emotion_dict[emotion_idx]["name"]
-                cv2.putText(frame, emotion_name, (x, y-10),
-                           cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0, 255, 0), 2)
-                
-                # Mettre à jour l'affichage de l'émotion et du message
-                with emotion_placeholder:
-                    st.markdown(f"""
-                    <div class="emotion-box">
-                        <div class="emotion-title">{emotion_name}</div>
-                    </div>
-                    """, unsafe_allow_html=True)
-                
-                with message_placeholder:
-                    st.markdown(f"""
-                    <div class="emotion-box">
-                        <div class="emotion-message">{emotion_dict[emotion_idx]["message"]}</div>
-                    </div>
-                    """, unsafe_allow_html=True)
-
-            # Afficher le flux vidéo
-            frame_placeholder.image(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
-            
-            # Petite pause pour éviter une surcharge
-            time.sleep(0.1)
-
-    except Exception as e:
-        st.error(f"Une erreur s'est produite : {str(e)}")
+    # Utiliser session_state pour partager les placeholders avec VideoProcessor
+    if 'emotion_placeholder' not in st.session_state:
+        st.session_state.emotion_placeholder = st.empty()
+    if 'message_placeholder' not in st.session_state:
+        st.session_state.message_placeholder = st.empty()
     
-    finally:
-        cap.release()
+    emotion_placeholder = st.session_state.emotion_placeholder
+    message_placeholder = st.session_state.message_placeholder
 
-else:
-    st.info("👆 Cliquez sur le bouton pour démarrer la détection d'émotions.")
+st.info("👆 Autorisez l'accès à la webcam dans votre navigateur pour démarrer la détection d'émotions.")