from flask import Flask, request, jsonify
import time
# import ktrain
# import gradio as gr

from fer import FER
import cv2
import gradio as gr
import json
import requests
import numpy as np

emotion_detector = FER(mtcnn=True)

def predict_image(url):

    # Convert Gradio image format (BGR) to RGB
    response = requests.get(url)
    img_array = np.frombuffer(response.content, np.uint8)
    img = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
    
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    analysis = emotion_detector.detect_emotions(img)
    start_time = time.time() 

    res = []
    for data in analysis:
      emotions = data['emotions']
      max_emotion = max(emotions, key=emotions.get)
      max_value = emotions[max_emotion]
      res.append({max_emotion: max_value})
  
    response = {
        'analysis': analysis,
        'best_predict': list(res),
    }
    return response


gr.Interface(fn=predict_image, inputs="textbox", outputs="json").launch()



# def predict_list():
#     data = request.json
#     messages = data.get('messages', [])
#     start_time = time.time() 
#     predictions = predictor.predict(messages)
    
#     response = {
#         'messages': messages,
#         'predictions': predictions,
#         'elapsed_time': time.time() - start_time
#     }
    
#     return jsonify(response)



# if __name__ == "__main__":
    # app.run(host="0.0.0.0", port=7860)