app.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from torchvision import transforms
import cv2
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):
    def __init__(self):
        super(EmotionCNN, self).__init__()
        self.conv_layers = nn.Sequential(
            nn.Conv2d(1, 32, 3, padding=1),
            nn.BatchNorm2d(32),
            nn.ReLU(),
            nn.Conv2d(32, 32, 3, padding=1),
            nn.BatchNorm2d(32),
            nn.ReLU(),
            nn.MaxPool2d(2),
            nn.Dropout2d(0.25),
            nn.Conv2d(32, 64, 3, padding=1),
            nn.BatchNorm2d(64),
            nn.ReLU(),
            nn.Conv2d(64, 64, 3, padding=1),
            nn.BatchNorm2d(64),
            nn.ReLU(),
            nn.MaxPool2d(2),
            nn.Dropout2d(0.25),
            nn.Conv2d(64, 128, 3, padding=1),
            nn.BatchNorm2d(128),
            nn.ReLU(),
            nn.Conv2d(128, 128, 3, padding=1),
            nn.BatchNorm2d(128),
            nn.ReLU(),
            nn.MaxPool2d(2),
            nn.Dropout2d(0.25)
        )
        self.fc_layers = nn.Sequential(
            nn.Linear(128 * 6 * 6, 512),
            nn.ReLU(),
            nn.Dropout(0.5),
            nn.Linear(512, 256),
            nn.ReLU(),
            nn.Dropout(0.5),
            nn.Linear(256, 8)
        )

    def forward(self, x):
        x = self.conv_layers(x)
        x = x.view(x.size(0), -1)
        x = self.fc_layers(x)
        return x

# Dictionnaire des émotions
emotion_dict = {
    0: {"name": "Colère", "message": "Respirez profondément et prenez un moment pour vous calmer."},
    1: {"name": "Mépris", "message": "Essayez de voir les choses d'un autre point de vue."},
    2: {"name": "Dégoût", "message": "Concentrez-vous sur les aspects positifs de la situation."},
    3: {"name": "Peur", "message": "Vous êtes en sécurité, prenez votre temps pour vous apaiser."},
    4: {"name": "Bonheur", "message": "Votre sourire illumine la pièce ! Continuez ainsi !"},
    5: {"name": "Neutre", "message": "Vous semblez calme et posé."},
    6: {"name": "Tristesse", "message": "Chaque jour est une nouvelle opportunité. Gardez espoir !"},
    7: {"name": "Surprise", "message": "La vie est pleine de surprises positives !"}
}

# Configuration de la page Streamlit
st.set_page_config(page_title="Détecteur d'Émotions", layout="wide")

# Styles CSS personnalisés
st.markdown("""
    <style>
    .main {
        background-color: #f5f5f5;
    }
    .stButton>button {
        background-color: #4CAF50;
        color: white;
        padding: 15px 32px;
        text-align: center;
        text-decoration: none;
        display: inline-block;
        font-size: 16px;
        margin: 4px 2px;
        cursor: pointer;
        border-radius: 12px;
        border: none;
        transition-duration: 0.4s;
    }
    .stButton>button:hover {
        background-color: #45a049;
    }
    .emotion-box {
        padding: 20px;
        border-radius: 10px;
        background-color: white;
        box-shadow: 0 4px 8px rgba(0,0,0,0.1);
        margin: 10px 0;
    }
    .emotion-title {
        color: #333;
        font-size: 24px;
        font-weight: bold;
        margin-bottom: 10px;
    }
    .emotion-message {
        color: #666;
        font-size: 18px;
        line-height: 1.5;
    }
    </style>
    """, unsafe_allow_html=True)

# Titre de l'application
st.title("🎭 Détecteur d'Émotions en Temps Réel")

# Initialisation du modèle
@st.cache_resource
def load_model():
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model = EmotionCNN().to(device)
    try:
        model.load_state_dict(torch.load("cnn_emotion_model.pth", map_location=device))
        model.eval()
    except Exception as e:
        st.error(f"Erreur lors du chargement du modèle : {str(e)}")
        st.stop()
    return model, device

# Chargement du modèle
model, device = load_model()

# Transformations pour l'image
transform = transforms.Compose([
    transforms.Grayscale(num_output_channels=1),
    transforms.Resize((48, 48)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.5], std=[0.5])
])

# Chargement du classificateur Haar pour la détection de visage
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
if face_cascade.empty():
    st.error("Erreur : Impossible de charger le classificateur Haar pour la détection de visage.")
    st.stop()

def detect_faces(frame):
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray, 1.1, 4)
    return faces

# Configuration RTC avec plusieurs STUN et TURN
RTC_CONFIGURATION = RTCConfiguration({
    "iceServers": [
        {"urls": "stun:stun.l.google.com:19302"},
        {"urls": "stun:stun1.l.google.com:19302"},
        {"urls": "stun:stun2.l.google.com:19302"},
        {"urls": "stun:stun3.l.google.com:19302"},
        {"urls": "stun:stun4.l.google.com:19302"},
        {"urls": "stun:stun.stunprotocol.org:3478"},
        # Exemple de configuration TURN (remplacez par vos propres identifiants si disponible)
        {
            "urls": "turn:your-turn-server.example.com:3478",
            "username": "your-username",
            "credential": "your-password"
        }
    ]
})

# 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):
        try:
            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
                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")
        except Exception as e:
            st.error(f"Erreur lors du traitement de la frame : {str(e)}")
            return frame

# Configuration de l'interface
col1, col2 = st.columns([2, 1])

with col1:
    st.markdown("### 📹 Flux Vidéo")
    try:
        webrtc_ctx = webrtc_streamer(
            key="emotion-detection",
            rtc_configuration=RTC_CONFIGURATION,
            video_processor_factory=VideoProcessor,
            media_stream_constraints={"video": True, "audio": False},
            async_processing=True
        )
    except Exception as e:
        st.error(f"Erreur lors de l'initialisation de WebRTC : {str(e)}")
        st.warning("Vérifiez votre connexion réseau ou les paramètres STUN/TURN.")

with col2:
    st.markdown("### 😊 Émotion Détectée")
    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()
    
    emotion_placeholder = st.session_state.emotion_placeholder
    message_placeholder = st.session_state.message_placeholder

st.info("👆 Autorisez l'accès à la webcam dans votre navigateur pour démarrer la détection d'émotions.")
st.warning("Si la connexion échoue, vérifiez votre réseau ou configurez un serveur TURN pour WebRTC.")

# Option de téléchargement d'image comme solution de secours
st.markdown("### 📷 Ou téléchargez une image")
uploaded_file = st.file_uploader("Choisissez une image...", type=["jpg", "jpeg", "png"])
if uploaded_file is not None:
    image = Image.open(uploaded_file)
    frame = np.array(image)
    faces = detect_faces(frame)
    
    for (x, y, w, h) in faces:
        cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 2)
        face_img = frame[y:y+h, x:x+w]
        pil_img = Image.fromarray(cv2.cvtColor(face_img, cv2.COLOR_BGR2RGB))
        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()
        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)
        
        emotion_placeholder.markdown(f"""
        <div class="emotion-box">
            <div class="emotion-title">{emotion_name}</div>
        </div>
        """, unsafe_allow_html=True)
        
        message_placeholder.markdown(f"""
        <div class="emotion-box">
            <div class="emotion-message">{emotion_dict[emotion_idx]["message"]}</div>
        </div>
        """, unsafe_allow_html=True)
    
    st.image(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))