import os
from fastai.vision.all import *
import gradio as gr

# Cargar los modelos
learn_emotion = load_learner('emotions_jey.pkl')
learn_emotion_labels = learn_emotion.dls.vocab

learn_sentiment = load_learner('sentiment_jey.pkl')
learn_sentiment_labels = learn_sentiment.dls.vocab

# Diccionario de mapeo de etiquetas en inglés a etiquetas en español
label_mapping = {
    'angry': 'enojado',
    'disgust': 'asco',
    'fear': 'miedo',
    'happy': 'feliz',
    'sad': 'triste',
    'surprise': 'sorpresa',
    'neutral': 'neutral',
    'negative': 'negativo',
    'positive': 'positivo'
}

# Función de predicción
def predict(img_path):
    img = PILImage.create(img_path)
    
    pred_emotion, pred_emotion_idx, probs_emotion = learn_emotion.predict(img)
    pred_sentiment, pred_sentiment_idx, probs_sentiment = learn_sentiment.predict(img)
    
    emotions = {label_mapping[label]: float(prob) for label, prob in zip(learn_emotion_labels, probs_emotion)}
    sentiments = {label_mapping[label]: float(prob) for label, prob in zip(learn_sentiment_labels, probs_sentiment)}
        
    return emotions, sentiments

# Interfaz de Gradio
title = "Detector de emociones y sentimientos faciales"
description = (
    "Esta interfaz utiliza redes neuronales para detectar emociones y sentimientos a partir de imágenes faciales."
)
article = "Esta herramienta proporciona una forma rápida de analizar emociones y sentimientos en imágenes."

examples = [
    'PrivateTest_10131363.jpg',
    'angry1.png',
    'angry2.jpg',
    'happy1.jpg',
    'happy2.jpg',
    'neutral1.jpg',
    'neutral2.jpg'
]

iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(shape=(48, 48), image_mode='L'),
    outputs=[gr.Label(label='Emoción'), gr.Label(label='Sentimiento')],
    title=title,
    examples=examples,
    description=description,
    article=article,
    allow_flagging='never'
)

iface.launch(enable_queue=True)






#################

import os
from fastai.vision.all import *
import gradio as gr

# Cargar los modelos
learn_emotion = load_learner('emotions_jey.pkl')
learn_emotion_labels = learn_emotion.dls.vocab

learn_sentiment = load_learner('sentiment_jey.pkl')
learn_sentiment_labels = learn_sentiment.dls.vocab

# Función de predicción
def predict(img_path):
    img = PILImage.create(img_path)
    
    pred_emotion, pred_emotion_idx, probs_emotion = learn_emotion.predict(img)
    pred_sentiment, pred_sentiment_idx, probs_sentiment = learn_sentiment.predict(img)
    
    emotions = {label: float(prob) for label, prob in zip(learn_emotion_labels, probs_emotion)}
    sentiments = {label: float(prob) for label, prob in zip(learn_sentiment_labels, probs_sentiment)}
        
    return emotions, sentiments

# Interfaz de Gradio
title = "Detector de emociones y sentimientos faciales "
description = (
    "Esta interfaz utiliza redes neuronales para detectar emociones y sentimientos a partir de imágenes faciales."
)
article = "Esta herramienta proporciona una forma rápida de analizar emociones y sentimientos en imágenes."

examples = [
    'PrivateTest_10131363.jpg',
    'angry1.png',
    'angry2.jpg',
    'happy1.jpg',
    'happy2.jpg',
    'neutral1.jpg',
    'neutral2.jpg'

]

iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(shape=(48, 48), image_mode='L'),
    outputs=[gr.Label(label='Emotion'), gr.Label(label='Sentiment')],
    title=title,
    examples=examples,
    description=description,
    article=article,
    allow_flagging='never'
)

iface.launch(enable_queue=True)