app.py

import google.generativeai as genai
from transformers import AutoFeatureExtractor, AutoModel
from PIL import Image
import torch
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Charger le modèle de vision par ordinateur
model_name = "google/efficientnet-b0"
feature_extractor = AutoFeatureExtractor.from_pretrained(model_name)
model = AutoModel.from_pretrained(model_name)

# Charger et prétraiter l'image
image_path = "path/to/your/image.jpg"
image = Image.open(image_path)
inputs = feature_extractor(images=image, return_tensors="pt")

# Passer l'image à travers le modèle
with torch.no_grad():
    outputs = model(**inputs)

# Extraire les caractéristiques de l'image
features = outputs.last_hidden_state.mean(dim=1)

# Charger le modèle de génération de texte
model_name = "t5-small"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Convertir les caractéristiques de l'image en texte
input_text = "Generate a recipe based on the following image features: " + str(features.tolist())
inputs = tokenizer(input_text, return_tensors="pt")

# Générer la recette
with torch.no_grad():
    outputs = model.generate(**inputs)

# Décoder la recette générée
recipe = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(recipe)