from PIL import Image
import numpy as np
import paddle.fluid as fluid
from 口罩检测.util import train_parameters
from 口罩检测.VGGNet import VGGNet
def load_image(img_path):
    img =Image.open(img_path)
    if img.mode !='RGB':
        img = img.covert('RGB')
    img = img.resize((244,244),Image.BILINEAR)
    img = np.array(img).astype('float32')
    img = img.transpose((2,0,1))
    img = img/255.0
    return img

label_dict = train_parameters['label_dict']

#模型预测
with fluid.dygraph.guard():
    model,_ = fluid.dygraph.load_dygraph('vgg')
    vgg = VGGNet()
    vgg.eval()
    infer_path='./unmask.jpg'
    img = Image.open(infer_path)

    x_data = load_image(infer_path)
    x_data = np.array(x_data)
    x_data = x_data[np.newaxis,:,:,:]
    x_data= fluid.dygraph.to_variable(x_data)
    out = vgg(x_data)
    result = np.argmax(out.numpy())
    print(label_dict)
    print("被预测的图片为：{}".format(label_dict[str(result)]))