tensorflow工程化测试TF-Hub库中的MobileNet_v2模型

tensorflow工程化测试TF-Hub库中的MobileNet_v2模型

加载分类模型hub.load_module_spec(“mobilenet_v2_100_224”)

from PIL import Image
from matplotlib import pyplot as plt
import numpy as np

import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()
import tensorflow_hub as hub
with open('中文标签.csv','rb') as f:
    f1=f.read().decode('gbk')
    labels=list(map(lambda x:x.replace(',',' '),list(f1)))
    print(len(labels),type(labels),labels[:5])
sample_images=['hy.jpg','ps.jpg','72.jpg']#定义待测图片路径
#加载分类模型
module_spec=hub.load_module_spec("mobilenet_v2_100_224")
#获取模型的输入图片尺寸
height,witdh=hub.get_expected_image_size(module_spec)
input_imgs=tf.placeholder(tf.float32,[None,height,witdh,3])#占位符
images=2*(input_imgs/255.0)-1.0#归一化
module=hub.Module(module_spec)#将模型载入张量图

logits=module(images)

y=tf.argmax(logits,axis=1)
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    sess.run(tf.tables_initializer())

    def preimg(img):    #图片预处理函数
        return np.asarray(img.resize((height,witdh)),dtype=np.float32).reshape(height,witdh,3)

    #获取原始图片与预处理图片
    batchImg=[preimg(Image.open(imgfilename)) for imgfilename in sample_images]
    orgImg=[Image.open(imgfilename) for imgfilename in sample_images]
    #样本输入模型
    # yv,img_norm=sess.run([y,images],feed_dict={input_imgs:batchImg})
    print(input_imgs)
    yv,img_norm=sess.run([y,images],feed_dict={input_imgs: batchImg})
    print(yv,np.shape(yv))

    def showresult(yy,img_norm,img_org):
        t1 = [img_org, (img_norm * 255).astype(np.uint8)]
        titles=["organizaton image","input image"]
        plt.figure()
        for i in range(2):
            plt.subplot(1,2,i+1)
            plt.imshow(t1[i])
            plt.axis('off')
            plt.title(titles[i])
        plt.show()
        print(yy,labels[yy])
    for yy,img1,img2 in zip(yv,batchImg,orgImg):
        showresult(yy,img1,img2)