import sys
sys.path.append('.')

from flask import Flask, request, jsonify
import os
import json
import cv2
import numpy as np

from tensorflow.keras.preprocessing import image
from keras_facenet import FaceNet
from sklearn.svm import OneClassSVM
import pickle

# create a facenet model
embedding_model = FaceNet(key = '20180402-114759',
    use_prebuilt=True,
    cache_folder= os.path.abspath(os.path.dirname(__file__)) + '/facenet_weights').model

clf_model = pickle.load(open(os.path.abspath(os.path.dirname(__file__)) + '/model/clf_mode.sav', 'rb'))

target_shape = (160, 160)

app = Flask(__name__) 

app.config['SITE'] = "http://0.0.0.0:8000/"
app.config['DEBUG'] = False

@app.route('/api/detect_human_face', methods=['POST'])
def detect_human_face():
    file1 = request.files['image1']
    image1 = cv2.imdecode(np.fromstring(file1.read(), np.uint8), cv2.IMREAD_COLOR)
    if image1 is None:
        result = "image1: is null!"
        status = "ok"
        response = jsonify({"status": status, "data": {"result": result}})
        response.status_code = 200
        response.headers["Content-Type"] = "application/json; charset=utf-8"
        return response

    X = np.float32([(np.float32(image1) - 127.5) / 127.5])
    X_ft = embedding_model.predict(X, batch_size=1)
    anomaly_score = clf_model.decision_function(X_ft) * -1

    if anomaly_score > 1:
        result = "Not Human"
    else:
        result = "Human"
  
    status = "ok"
    response = jsonify(
    {
        "status": status, 
        "data": {
            "result": result, 
            "anomaly_score": float(anomaly_score)
            }
    })

    response.status_code = 200
    response.headers["Content-Type"] = "application/json; charset=utf-8"
    return response

if __name__ == '__main__':
    port = int(os.environ.get("PORT", 8000))
    app.run(host='0.0.0.0', port=port)