图片识别代码问题

小鸭不学习

创建了四个独立的.py文件合起来实现手写数字识别


有没有大佬能帮着看一下下面的问题，因为1.x版本的Tensorflow已经下架，但是这里面很多编码都是1.X版本的Tensorflow，所以可能导致错误，可以帮着改一下不，就是不要改变大致目标

第一个文件：mnist_forward.py
-源代码：

import tensorflow as tf

INPUT_NODE = 784
OUTPUT_NODE = 10
LAYER1_NODE = 500

def get_weight(shape, regularizer):
    w = tf.Variable(tf.random.truncated_normal(shape, stddev=0.1))
    if regularizer != None: 
        tf.add_to_collection("losses", tf.keras.regularizers.l2(regularizer)(w))
    return w

def get_bias(shape):
    b = tf.Variable(tf.zeros(shape))
    return b

def forward(x, regularizer):
    w1 = get_weight([INPUT_NODE, LAYER1_NODE], regularizer)
    b1 = get_bias([LAYER1_NODE])
    y1 = tf.nn.relu(tf.matmul(x, w1) + b1)

    w2 = get_weight([LAYER1_NODE, OUTPUT_NODE], regularizer)
    b2 = get_bias([OUTPUT_NODE])
    y = tf.matmul(y1, w2) + b2
    return y

运行结果：无报错信息，无任何执行结果，执行后一段时间自动跳出

第二个文件：mnist_backward.py
-源代码：

import tensorflow as tf
from tensorflow.keras.datasets import mnist
from mnist_forward import INPUT_NODE, OUTPUT_NODE  # 修正导入错误
import os

BATCH_SIZE = 200
LEARNING_RATE_BASE = 0.1
LEARNING_RATE_DECAY = 0.99
REGULARIZER = 0.0001
STEPS = 50000
MOVING_AVERAGE_DECAY = 0.99
MODEL_SAVE_PATH = "./model/"
MODEL_NAME = "mnist_model"

def backward(dataset):

    x = tf.placeholder(tf.float32, [None, INPUT_NODE])  # 使用修正后的导入
    y_ = tf.placeholder(tf.float32, [None, OUTPUT_NODE])  # 使用修正后的导入
    y = mnist_forward.forward(x, REGULARIZER)
    global_step = tf.Variable(0, trainable=False)

    logits = y  # 将神经网络输出 y 作为 logits
    labels = tf.argmax(y_, 1)  # 将类别索引作为 labels
    ce = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, labels=labels)  # 修改损失计算方式
    cem = tf.reduce_mean(ce)
    loss = cem + tf.add_n(tf.get_collection("losses"))

    learning_rate = tf.train.exponential_decay(
        LEARNING_RATE_BASE,
        global_step,
        dataset.train.num_examples / BATCH_SIZE,
        LEARNING_RATE_DECAY,
        staircase=True)

    train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)

    ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
    ema_op = ema.apply(tf.trainable_variables())
    with tf.control_dependencies([train_step, ema_op]):
        train_op = tf.no_op(name="train")

    saver = tf.train.Saver()

    with tf.Session() as sess:
        init_op = tf.global_variables_initializer()
        sess.run(init_op)

        ckpt = tf.train.get_checkpoint_state(MODEL_SAVE_PATH)
        if ckpt and ckpt.model_checkpoint_path:
            saver.restore(sess, ckpt.model_checkpoint_path)

        for i in range(STEPS):
            xs, ys = dataset.train.next_batch(BATCH_SIZE)
            _, lossvalue, step = sess.run([train_op, loss, global_step], feed_dict={x: xs, y_: ys})
            if i % 1000 == 0:
                print("After %d training step(s), loss on training batch is %g." % (step, lossvalue))
                saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step=global_step)


def main():
    (x_train, y_train), _ = mnist.load_data()
    x_train = x_train.reshape(x_train.shape[0], -1) / 255.0
    y_train = tf.keras.utils.to_categorical(y_train, 10)
    dataset = tf.data.Dataset.from_tensor_slices((x_train, y_train)).shuffle(10000).batch(BATCH_SIZE)
    backward(dataset)

if __name__ == "__main__":
    main()

-运行结果：无报错信息，执行以后只要你点了IDLE的弹窗口就会卡死然后你必须手动退出，如果你让它往下执行，它就会出现很多字符串，想下载器一样，时不时会卡死一段时间，然后接着执行，运行大概五分钟以后，就会彻底卡死，需要手动退出，下面给一个运行截图：

                               
登录/注册后可看大图

第三个文件：mnist_test.py
-源代码：

import tensorflow as tf
import numpy as np
import mnist_forward
import os

TEST_INTERVAL_SECS = 5

def test(mnist):
    x = tf.placeholder(tf.float32, [None, mnist_forward.INPUT_NODE])
    y_ = tf.placeholder(tf.float32, [None, mnist_forward.OUTPUT_NODE])
    y = mnist_forward.forward(x, None)

    ema = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
    ema_restore = ema.variables_to_restore()
    saver = tf.train.Saver(ema_restore)

    correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

    while True:
        latest_checkpoint = tf.train.latest_checkpoint(mnist_backward.MODEL_SAVE_PATH)
        if latest_checkpoint:
            saver.restore(tf.compat.v1.get_default_session(), latest_checkpoint)
            global_step = int(latest_checkpoint.split("-")[-1])
            accuracy_score = accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels})
            print("After %s training step(s), test accuracy = %g" % (global_step, accuracy_score))
        else:
            print("No checkpoint file found")
            return
        time.sleep(TEST_INTERVAL_SECS)

def main():
    mnist = tf.keras.datasets.mnist
    (x_train, y_train), (x_test, y_test) = mnist.load_data()
    x_test = np.reshape(x_test, (-1, mnist_forward.INPUT_NODE))
    y_test = tf.keras.utils.to_categorical(y_test, mnist_forward.OUTPUT_NODE)
    mnist.test = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(10000)
    test(mnist)

if __name__ == "__main__":
    main()

-运行结果：
同样没有报错信息，但是结果跟mnist.backward.py的结果一样，非常卡顿，运行一段时间后会自动退出，出现的数据也是一抹一样

第四个文件：mnist_app.py
-源代码：

import tensorflow.compat.v1 as tf
import numpy as np
from PIL import Image
import mnist_backward
import mnist_forward
import os

tf.compat.v1.disable_v2_behavior()

def restore_model(testPicArr):
    with tf.Graph().as_default() as tg:
        x = tf.compat.v1.placeholder(tf.float32, [None, mnist_forward.INPUT_NODE])
        y = mnist_forward.forward(x, None)
        preValue = tf.argmax(y, 1)

        variable_averages = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.compat.v1.train.Saver(variables_to_restore)

        with tf.compat.v1.Session() as sess:
            ckpt = tf.train.get_checkpoint_state(mnist_backward.MODEL_SAVE_PATH)
            if ckpt and ckpt.model_checkpoint_path:
                print("Checkpoint file path:", ckpt.model_checkpoint_path)
                saver.restore(sess, ckpt.model_checkpoint_path)

                preValue = sess.run(preValue, feed_dict={x: testPicArr})
                return preValue
            else:
                print("No checkpoint file found in", mnist_backward.MODEL_SAVE_PATH)
                return -1

def pre_pic(picName):
    img = Image.open(picName)
    reIm = img.resize((28, 28), Image.LANCZOS)
    im_arr = np.array(reIm.convert('L'))
    threshold = 50
    for i in range(28):
        for j in range(28):
            im_arr[i][j] = 255 - im_arr[i][j]
            if im_arr[i][j] < threshold:
                im_arr[i][j] = 0
            else:
                im_arr[i][j] = 255
    nm_arr = im_arr.reshape([1, 784])
    nm_arr = nm_arr.astype(np.float32)
    img_ready = np.multiply(nm_arr, 1.0 / 255.0)
    return img_ready

def application():
    testNum = int(input("input the number of test pictures:"))
    for i in range(testNum):
        testPic = input("the path of test picture:")
        testPicArr = pre_pic(testPic)
        preValue = restore_model(testPicArr)
        print("The prediction number is:", preValue)

def main():
    application()

if __name__ == "__main__":
    main()

-运行结果：
（1）警告信息：

WARNING:tensorflow:From D:\卢天恩文件夹\3.编程\0.Tendsorflow学习\2.代码实践\手写数字识别\源代码\mnist_app.py:8: The name tf.disable_v2_behavior is deprecated. Please use tf.compat.v1.disable_v2_behavior instead.

WARNING:tensorflow:From C:\Users\Administrator\AppData\Local\Programs\Python\Python311\Lib\site-packages\tensorflow\python\compat\v2_compat.py:98: disable_resource_variables (from tensorflow.python.ops.resource_variables_toggle) is deprecated and will be removed in a future version.
Instructions for updating:
non-resource variables are not supported in the long term

（2）运行时无法查找节点，无法识别图片：

input the number of test pictures:3
the path of test picture:pic/3.jpg
WARNING:tensorflow:From D:\卢天恩文件夹\3.编程\0.Tendsorflow学习\2.代码实践\手写数字识别\源代码\mnist_app.py:18: The name tf.train.Saver is deprecated. Please use tf.compat.v1.train.Saver instead.

No checkpoint file found in ./model/
The prediction number is: -1
the path of test picture:pic/10.jpg
No checkpoint file found in ./model/
The prediction number is: -1
the path of test picture:pic/4.jpg
No checkpoint file found in ./model/
The prediction number is: -1

请求帮我看一下，可能是这四个文件都有问题，都互相关联，向前传播的过程没报错没结果，一段时间就退出了，向后传播和优化训练都会卡死

FishC

问题可能出在两个地方：

1. TensorFlow版本不兼容：
首先，你提到的这些代码是基于TensorFlow 1.x版本编写的，而你可能在使用TensorFlow 2.x版本。在TensorFlow 2.x版本中，一些API和函数名称发生了变化，所以代码可能无法正常运行。

解决方法：
将`import tensorflow.compat.v1 as tf`替换为`import tensorflow as tf`，然后将所有使用`tf.compat.v1`前缀的函数和类名改为使用`tf`即可。

2. 模型文件缺失或路径错误：
在`mnist_app.py`文件中，你尝试从模型文件中恢复模型权重以进行预测。可能是由于模型文件丢失或路径设置不正确，导致无法找到模型文件。

解决方法：
确保模型文件`.ckpt`位于指定的模型保存路径`MODEL_SAVE_PATH`下，并且命名为`mnist_model.ckpt`。另外，请确保`mnist_test.py`文件中的`TEST_INTERVAL_SECS`变量设置的时间足够长，以允许模型训练完成并生成模型文件。

针对以上问题的解决方法，你可以按照以下步骤进行：

步骤1：修改`mnist_forward.py`文件
将代码中的`get_weight`和`get_bias`函数的实现修改如下：

def get_weight(shape, regularizer):
    w = tf.Variable(tf.random.truncated_normal(shape, stddev=0.1))
    if regularizer != None:
        tf.add_to_collection("losses", tf.keras.regularizers.l2(regularizer)(w))
    return w

def get_bias(shape):
    b = tf.Variable(tf.zeros(shape))
    return b

步骤2：修改`mnist_test.py`文件
将代码中的`test`函数和`main`函数的实现修改如下：

def test(mnist):
    x = tf.placeholder(tf.float32, shape=(None, mnist_forward.INPUT_NODE))
    y_ = tf.placeholder(tf.float32, shape=(None, mnist_forward.OUTPUT_NODE))
    y = mnist_forward.forward(x, None)

    ema = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
    ema_restore = ema.variables_to_restore()
    saver = tf.train.Saver(ema_restore)

    correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

    while True:
        latest_checkpoint = tf.train.latest_checkpoint(mnist_backward.MODEL_SAVE_PATH)
        if latest_checkpoint:
            with tf.compat.v1.Session() as sess:
                saver.restore(sess, latest_checkpoint)
                global_step = int(latest_checkpoint.split("-")[-1])
                accuracy_score = sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels})
                print("After %d training step(s), test accuracy = %g" % (global_step, accuracy_score))
        else:
            print("No checkpoint file found")
            return

def main():
    mnist = tf.keras.datasets.mnist
    (x_train, y_train), (x_test, y_test) = mnist.load_data()
    x_test = x_test.reshape(-1, mnist_forward.INPUT_NODE) / 255.0
    y_test = tf.keras.utils.to_categorical(y_test, mnist_forward.OUTPUT_NODE)
    mnist.test = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(10000)
    test(mnist)

if __name__ == "__main__":
    main()

步骤3：修改`mnist_app.py`文件
将代码中的`restore_model`函数和`pre_pic`函数的实现修改如下：

def restore_model(testPicArr):
    with tf.Graph().as_default() as g:
        x = tf.placeholder(tf.float32, shape=(None, mnist_forward.INPUT_NODE))
        y = mnist_forward.forward(x, None)
        preValue = tf.argmax(y, 1)

        variable_averages = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.train.Saver(variables_to_restore)

        with tf.compat.v1.Session(graph=g) as sess:
            latest_checkpoint = tf.train.latest_checkpoint(mnist_backward.MODEL_SAVE_PATH)
            if latest_checkpoint:
                saver.restore(sess, latest_checkpoint)
                preValue = sess.run(preValue, feed_dict={x: [testPicArr]})
                return preValue[0]
            else:
                print("No checkpoint file found in", mnist_backward.MODEL_SAVE_PATH)
                return -1

def pre_pic(picName):
    img = Image.open(picName).convert('L')
    reIm = img.resize((28, 28), Image.LANCZOS)
    im_arr = np.array(reIm)
    threshold = 50
    im_arr = np.where(im_arr < threshold, 0, 255)
    nm_arr = im_arr.reshape(1, -1).astype(np.float32)
    img_ready = nm_arr / 255.0
    return img_ready

def application():
    testNum = int(input("input the number of test pictures:"))
    for i in range(testNum):
        testPic = input("the path of test picture:")
        testPicArr = pre_pic(testPic)
        preValue = restore_model(testPicArr)
        print("The prediction number is:", preValue)

def main():
    application()

if __name__ == "__main__":
    main()

最后，确保所有的模型文件和图片路径都正确设置，然后尝试运行代码。希望这些修改能够解决你的问题！如果还有其他问题，请随时提问。

以上回复来自 FishC AI 小助理，如未能正确解答您的问题，请继续追问。