wound_predict

# -*- coding: utf-8 -*-
"""
20190109
Created on Wed Dec 19 15:23:18 2018

@author: adi
"""
from keras.models import load_model
from keras.preprocessing import image
import numpy as np
import cv2

def pre():
    classes=['擦傷','瘀青','割裂傷','抓傷']
    cap = cv2.VideoCapture(0)
    print('======== Load Model =========')
    model=load_model("wound0103MOBILENET.h5")
    print('======== Load Done =========')
    #img = image.load_img("image 1.jpg", target_size=(224, 224))
    '''img=cv2.imread("1.jpeg")
    img=cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    img.resize(224,224,3)'''
    while True:
        ret, img = cap.read()
        img=cv2.resize(img,(224,224))
        #img=cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        img_tensor = image.img_to_array(img)                    # (height, width, channels)
        img_tensor=img_tensor/255.
        img_tensor = np.expand_dims(img_tensor, axis=0)  
        pred=model.predict(img_tensor)
        cv2.imshow('Video_FR', img)
        if cv2.waitKey(1)&0xFF==ord('q'):
            cv2.destroyAllWindows()
            break
        print(pred)
        if np.max(pred)>0.6:
            cv2.destroyAllWindows()
            break
    return classes[np.argmax(pred)]     
'''
        img_tensor = image.img_to_array(img)                 # (height, width, channels)
        img_tensor=img_tensor/255.
        #print(img_tensor)
        img_tensor = np.expand_dims(img_tensor, axis=0)  
        pred=model.predict(img_tensor)
        print(pred)
        
        print(classes[np.argmax(pred)])
        '''