Tensorflow Inception resnet v2 输入张量

Question

我正在尝试运行此代码

import os
import tensorflow as tf
from datasets import imagenet
from nets import inception_resnet_v2
from preprocessing import inception_preprocessing

checkpoints_dir = 'model'

slim = tf.contrib.slim

batch_size = 3
image_size = 299

with tf.Graph().as_default():

with slim.arg_scope(inception_resnet_v2.inception_resnet_v2_arg_scope()):
    logits, _ = inception_resnet_v2.inception_resnet_v2([1, 299, 299, 3], num_classes=1001, is_training=False)
    probabilities = tf.nn.softmax(logits)

    init_fn = slim.assign_from_checkpoint_fn(
    os.path.join(checkpoints_dir, 'inception_resnet_v2_2016_08_30.ckpt'),
    slim.get_model_variables('InceptionResnetV2'))

    with tf.Session() as sess:
        init_fn(sess)

        imgPath = '.../image_3.jpeg'
        testImage_string = tf.gfile.FastGFile(imgPath, 'rb').read()
        testImage = tf.image.decode_jpeg(testImage_string, channels=3)

        np_image, probabilities = sess.run([testImage, probabilities])
        probabilities = probabilities[0, 0:]
        sorted_inds = [i[0] for i in sorted(enumerate(-probabilities), key=lambda x:x[1])]

        names = imagenet.create_readable_names_for_imagenet_labels()
        for i in range(15):
            index = sorted_inds[i]
            print((probabilities[index], names[index]))

但是TF显示错误：ValueError: rank of shape must be at least 4 not: 1

我认为问题在于输入张量形状[1, 299, 299, 3]。如何为 3 通道 JPEG 图像输入张量？？？

还有一个类似的问题 (Using pre-trained inception_resnet_v2 with Tensorflow)。我在代码 input_tensor 中看到了 - 不幸的是，有解释什么是 input_tensor。也许我在问一些不言而喻的事情，但我卡住了！非常感谢您的任何建议！

Answer 1

您必须对图像进行预处理。这是一个代码：

import os
import tensorflow as tf
from datasets import imagenet
from nets import inception_resnet_v2
from preprocessing import inception_preprocessing

checkpoints_dir = 'model'

slim = tf.contrib.slim

batch_size = 3
image_size = 299

with tf.Graph().as_default():
    with slim.arg_scope(inception_resnet_v2.inception_resnet_v2_arg_scope()):

        imgPath = '.../cat.jpg'
        testImage_string = tf.gfile.FastGFile(imgPath, 'rb').read()
        testImage = tf.image.decode_jpeg(testImage_string, channels=3)
        processed_image = inception_preprocessing.preprocess_image(testImage, image_size, image_size, is_training=False)
        processed_images = tf.expand_dims(processed_image, 0)

        logits, _ = inception_resnet_v2.inception_resnet_v2(processed_images, num_classes=1001, is_training=False)
        probabilities = tf.nn.softmax(logits)

        init_fn = slim.assign_from_checkpoint_fn(
        os.path.join(checkpoints_dir, 'inception_resnet_v2_2016_08_30.ckpt'), slim.get_model_variables('InceptionResnetV2'))

        with tf.Session() as sess:
            init_fn(sess)

            np_image, probabilities = sess.run([processed_images, probabilities])
            probabilities = probabilities[0, 0:]
            sorted_inds = [i[0] for i in sorted(enumerate(-probabilities), key=lambda x: x[1])]

            names = imagenet.create_readable_names_for_imagenet_labels()
            for i in range(15):
                index = sorted_inds[i]
                print((probabilities[index], names[index]))

答案是：

(0.1131034, 'tiger cat')
(0.079478227, 'tabby, tabby cat')
(0.052777905, 'Cardigan, Cardigan Welsh corgi')
(0.030195976, 'laptop, laptop computer')
(0.027841948, 'bathtub, bathing tub, bath, tub')
(0.026694898, 'television, television system')
(0.024981709, 'carton')
(0.024039172, 'Egyptian cat')
(0.018425584, 'tub, vat')
(0.018221909, 'Pembroke, Pembroke Welsh corgi')
(0.015066789, 'skunk, polecat, wood pussy')
(0.01377619, 'screen, CRT screen')
(0.012509955, 'monitor')
(0.012224807, 'mouse, computer mouse')
(0.012188354, 'refrigerator, icebox')

Answer 2

您可以使用tf.expand_dims(your_tensor_3channel, axis=0) 将其扩展为批处理格式。