app.py

import gradio as gr
import torch
from transformers import (
    BlipProcessor, 
    BlipForQuestionAnswering,
    pipeline,
    AutoTokenizer, 
    AutoModelForCausalLM
)
from PIL import Image
import os
import logging

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

class MultimodalProcessor:
    def __init__(self):
        self.load_models()
    
    def load_models(self):
        """Charge les modèles avec gestion d'erreurs"""
        try:
            logger.info("Chargement des modèles...")
            self.blip_processor = BlipProcessor.from_pretrained("Salesforce/blip-vqa-base")
            self.blip_model = BlipForQuestionAnswering.from_pretrained("Salesforce/blip-vqa-base")
            self.audio_transcriber = pipeline("automatic-speech-recognition", 
                                           model="openai/whisper-small")
            self.text_generator = pipeline("text-generation", 
                                        model="gpt2")
            logger.info("Modèles chargés avec succès")
        except Exception as e:
            logger.error(f"Erreur lors du chargement des modèles: {str(e)}")
            raise

    def analyze_image(self, image):
        """Analyse une image et retourne une description"""
        try:
            if image is None:
                return ""
            
            questions = [
                "What is in the picture?",
                "What are the main colors?",
                "What is the setting or background?",
                "What is happening in the image?",
            ]
            
            responses = {}
            for question in questions:
                inputs = self.blip_processor(images=image, text=question, return_tensors="pt")
                outputs = self.blip_model.generate(**inputs)
                answer = self.blip_processor.decode(outputs[0], skip_special_tokens=True)
                responses[question] = answer
            
            description = (
                f"This image shows {responses['What is in the picture?']}. "
                f"The main colors are {responses['What are the main colors?']}. "
                f"The setting is {responses['What is the setting or background?']}. "
                f"In the scene, {responses['What is happening in the image?']}"
            )
            
            return description
        except Exception as e:
            logger.error(f"Erreur lors de l'analyse de l'image: {str(e)}")
            return "Erreur lors de l'analyse de l'image."

    def transcribe_audio(self, audio_path):
        """Transcrit un fichier audio"""
        try:
            if audio_path is None:
                return ""
            return self.audio_transcriber(audio_path)["text"]
        except Exception as e:
            logger.error(f"Erreur lors de la transcription audio: {str(e)}")
            return "Erreur lors de la transcription audio."

    def generate_text(self, prompt):
        """Génère du texte à partir d'un prompt"""
        try:
            if not prompt:
                return ""
            response = self.text_generator(prompt, 
                                        max_length=200,
                                        num_return_sequences=1)[0]["generated_text"]
            return response
        except Exception as e:
            logger.error(f"Erreur lors de la génération de texte: {str(e)}")
            return "Erreur lors de la génération de texte."

    def process_inputs(self, image, audio, text):
        """Traite les entrées multimodales"""
        try:
            # Analyse de l'image
            image_description = self.analyze_image(image) if image is not None else ""
            
            # Transcription audio
            audio_text = self.transcribe_audio(audio) if audio is not None else ""
            
            # Combinaison des entrées
            combined_input = ""
            if image_description:
                combined_input += f"Visual description: {image_description}\n"
            if audio_text:
                combined_input += f"Audio content: {audio_text}\n"
            if text:
                combined_input += f"Additional context: {text}\n"
            
            # Génération du prompt final
            if combined_input:
                final_prompt = self.generate_text(combined_input)
            else:
                final_prompt = "Aucune entrée fournie."
            
            return final_prompt
            
        except Exception as e:
            logger.error(f"Erreur lors du traitement des entrées: {str(e)}")
            return "Une erreur est survenue lors du traitement des entrées."

def create_interface():
    """Crée l'interface Gradio"""
    processor = MultimodalProcessor()
    
    interface = gr.Interface(
        fn=processor.process_inputs,
        inputs=[
            gr.Image(type="pil", label="Télécharger une image"),
            gr.Audio(type="filepath", label="Télécharger un fichier audio"),
            gr.Textbox(label="Entrez du texte additionnel")
        ],
        outputs=[
            gr.Textbox(label="Description générée")
        ],
        title="Analyseur de Contenu Multimodal",
        description="""
        Cette application analyse vos contenus multimodaux :
        - Images : génère une description détaillée
        - Audio : transcrit le contenu
        - Texte : enrichit la description
        
        La sortie combine toutes ces informations en une description cohérente.
        """
    )
    
    return interface

if __name__ == "__main__":
    interface = create_interface()
    interface.launch()