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：

1. import tensorflow as tf
   
2. 
   
3. INPUT_NODE = 784
   
4. OUTPUT_NODE = 10
   
5. LAYER1_NODE = 500
   
6. 
   
7. def get_weight(shape, regularizer):
   
8.     w = tf.Variable(tf.random.truncated_normal(shape, stddev=0.1))
   
9.     if regularizer != None:
   
10.         tf.add_to_collection("losses", tf.keras.regularizers.l2(regularizer)(w))
    
11.     return w
    
12. 
    
13. def get_bias(shape):
    
14.     b = tf.Variable(tf.zeros(shape))
    
15.     return b
    
16. 
    
17. def forward(x, regularizer):
    
18.     w1 = get_weight([INPUT_NODE, LAYER1_NODE], regularizer)
    
19.     b1 = get_bias([LAYER1_NODE])
    
20.     y1 = tf.nn.relu(tf.matmul(x, w1) + b1)
    
21. 
    
22.     w2 = get_weight([LAYER1_NODE, OUTPUT_NODE], regularizer)
    
23.     b2 = get_bias([OUTPUT_NODE])
    
24.     y = tf.matmul(y1, w2) + b2
    
25.     return y
    

复制代码

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：

1. import tensorflow as tf
   
2. from tensorflow.keras.datasets import mnist
   
3. from mnist_forward import INPUT_NODE, OUTPUT_NODE  # 修正导入错误
   
4. import os
   
5. 
   
6. BATCH_SIZE = 200
   
7. LEARNING_RATE_BASE = 0.1
   
8. LEARNING_RATE_DECAY = 0.99
   
9. REGULARIZER = 0.0001
   
10. STEPS = 50000
    
11. MOVING_AVERAGE_DECAY = 0.99
    
12. MODEL_SAVE_PATH = "./model/"
    
13. MODEL_NAME = "mnist_model"
    
14. 
    
15. def backward(dataset):
    
16. 
    
17.     x = tf.placeholder(tf.float32, [None, INPUT_NODE])  # 使用修正后的导入
    
18.     y_ = tf.placeholder(tf.float32, [None, OUTPUT_NODE])  # 使用修正后的导入
    
19.     y = mnist_forward.forward(x, REGULARIZER)
    
20.     global_step = tf.Variable(0, trainable=False)
    
21. 
    
22.     logits = y  # 将神经网络输出 y 作为 logits
    
23.     labels = tf.argmax(y_, 1)  # 将类别索引作为 labels
    
24.     ce = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, labels=labels)  # 修改损失计算方式
    
25.     cem = tf.reduce_mean(ce)
    
26.     loss = cem + tf.add_n(tf.get_collection("losses"))
    
27. 
    
28.     learning_rate = tf.train.exponential_decay(
    
29.         LEARNING_RATE_BASE,
    
30.         global_step,
    
31.         dataset.train.num_examples / BATCH_SIZE,
    
32.         LEARNING_RATE_DECAY,
    
33.         staircase=True)
    
34. 
    
35.     train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss, global_step=global_step)
    
36. 
    
37.     ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
    
38.     ema_op = ema.apply(tf.trainable_variables())
    
39.     with tf.control_dependencies([train_step, ema_op]):
    
40.         train_op = tf.no_op(name="train")
    
41. 
    
42.     saver = tf.train.Saver()
    
43. 
    
44.     with tf.Session() as sess:
    
45.         init_op = tf.global_variables_initializer()
    
46.         sess.run(init_op)
    
47. 
    
48.         ckpt = tf.train.get_checkpoint_state(MODEL_SAVE_PATH)
    
49.         if ckpt and ckpt.model_checkpoint_path:
    
50.             saver.restore(sess, ckpt.model_checkpoint_path)
    
51. 
    
52.         for i in range(STEPS):
    
53.             xs, ys = dataset.train.next_batch(BATCH_SIZE)
    
54.             _, lossvalue, step = sess.run([train_op, loss, global_step], feed_dict={x: xs, y_: ys})
    
55.             if i % 1000 == 0:
    
56.                 print("After %d training step(s), loss on training batch is %g." % (step, lossvalue))
    
57.                 saver.save(sess, os.path.join(MODEL_SAVE_PATH, MODEL_NAME), global_step=global_step)
    
58. 
    
59. 
    
60. def main():
    
61.     (x_train, y_train), _ = mnist.load_data()
    
62.     x_train = x_train.reshape(x_train.shape[0], -1) / 255.0
    
63.     y_train = tf.keras.utils.to_categorical(y_train, 10)
    
64.     dataset = tf.data.Dataset.from_tensor_slices((x_train, y_train)).shuffle(10000).batch(BATCH_SIZE)
    
65.     backward(dataset)
    
66. 
    
67. if __name__ == "__main__":
    
68.     main()
    

复制代码

-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：

1. import tensorflow as tf
   
2. import numpy as np
   
3. import mnist_forward
   
4. import os
   
5. 
   
6. TEST_INTERVAL_SECS = 5
   
7. 
   
8. def test(mnist):
   
9.     x = tf.placeholder(tf.float32, [None, mnist_forward.INPUT_NODE])
   
10.     y_ = tf.placeholder(tf.float32, [None, mnist_forward.OUTPUT_NODE])
    
11.     y = mnist_forward.forward(x, None)
    
12. 
    
13.     ema = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
    
14.     ema_restore = ema.variables_to_restore()
    
15.     saver = tf.train.Saver(ema_restore)
    
16. 
    
17.     correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
    
18.     accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    
19. 
    
20.     while True:
    
21.         latest_checkpoint = tf.train.latest_checkpoint(mnist_backward.MODEL_SAVE_PATH)
    
22.         if latest_checkpoint:
    
23.             saver.restore(tf.compat.v1.get_default_session(), latest_checkpoint)
    
24.             global_step = int(latest_checkpoint.split("-")[-1])
    
25.             accuracy_score = accuracy.eval(feed_dict={x: mnist.test.images, y_: mnist.test.labels})
    
26.             print("After %s training step(s), test accuracy = %g" % (global_step, accuracy_score))
    
27.         else:
    
28.             print("No checkpoint file found")
    
29.             return
    
30.         time.sleep(TEST_INTERVAL_SECS)
    
31. 
    
32. def main():
    
33.     mnist = tf.keras.datasets.mnist
    
34.     (x_train, y_train), (x_test, y_test) = mnist.load_data()
    
35.     x_test = np.reshape(x_test, (-1, mnist_forward.INPUT_NODE))
    
36.     y_test = tf.keras.utils.to_categorical(y_test, mnist_forward.OUTPUT_NODE)
    
37.     mnist.test = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(10000)
    
38.     test(mnist)
    
39. 
    
40. if __name__ == "__main__":
    
41.     main()
    
42. 
    

复制代码

-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：

1. import tensorflow.compat.v1 as tf
   
2. import numpy as np
   
3. from PIL import Image
   
4. import mnist_backward
   
5. import mnist_forward
   
6. import os
   
7. 
   
8. tf.compat.v1.disable_v2_behavior()
   
9. 
   
10. def restore_model(testPicArr):
    
11.     with tf.Graph().as_default() as tg:
    
12.         x = tf.compat.v1.placeholder(tf.float32, [None, mnist_forward.INPUT_NODE])
    
13.         y = mnist_forward.forward(x, None)
    
14.         preValue = tf.argmax(y, 1)
    
15. 
    
16.         variable_averages = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
    
17.         variables_to_restore = variable_averages.variables_to_restore()
    
18.         saver = tf.compat.v1.train.Saver(variables_to_restore)
    
19. 
    
20.         with tf.compat.v1.Session() as sess:
    
21.             ckpt = tf.train.get_checkpoint_state(mnist_backward.MODEL_SAVE_PATH)
    
22.             if ckpt and ckpt.model_checkpoint_path:
    
23.                 print("Checkpoint file path:", ckpt.model_checkpoint_path)
    
24.                 saver.restore(sess, ckpt.model_checkpoint_path)
    
25. 
    
26.                 preValue = sess.run(preValue, feed_dict={x: testPicArr})
    
27.                 return preValue
    
28.             else:
    
29.                 print("No checkpoint file found in", mnist_backward.MODEL_SAVE_PATH)
    
30.                 return -1
    
31. 
    
32. def pre_pic(picName):
    
33.     img = Image.open(picName)
    
34.     reIm = img.resize((28, 28), Image.LANCZOS)
    
35.     im_arr = np.array(reIm.convert('L'))
    
36.     threshold = 50
    
37.     for i in range(28):
    
38.         for j in range(28):
    
39.             im_arr[i][j] = 255 - im_arr[i][j]
    
40.             if im_arr[i][j] < threshold:
    
41.                 im_arr[i][j] = 0
    
42.             else:
    
43.                 im_arr[i][j] = 255
    
44.     nm_arr = im_arr.reshape([1, 784])
    
45.     nm_arr = nm_arr.astype(np.float32)
    
46.     img_ready = np.multiply(nm_arr, 1.0 / 255.0)
    
47.     return img_ready
    
48. 
    
49. def application():
    
50.     testNum = int(input("input the number of test pictures:"))
    
51.     for i in range(testNum):
    
52.         testPic = input("the path of test picture:")
    
53.         testPicArr = pre_pic(testPic)
    
54.         preValue = restore_model(testPicArr)
    
55.         print("The prediction number is:", preValue)
    
56. 
    
57. def main():
    
58.     application()
    
59. 
    
60. if __name__ == "__main__":
    
61.     main()
    
62. 
    

复制代码

-运行结果：
（1）警告信息：

1. 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.
   
2. 
   
3. 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.
   
4. Instructions for updating:
   
5. non-resource variables are not supported in the long term

复制代码

（2）运行时无法查找节点，无法识别图片：

1. input the number of test pictures:3
   
2. the path of test picture:pic/3.jpg
   
3. 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.
   
4. 
   
5. No checkpoint file found in ./model/
   
6. The prediction number is: -1
   
7. the path of test picture:pic/10.jpg
   
8. No checkpoint file found in ./model/
   
9. The prediction number is: -1
   
10. the path of test picture:pic/4.jpg
    
11. No checkpoint file found in ./model/
    
12. The prediction number is: -1

复制代码

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：

1. def get_weight(shape, regularizer):
   
2.     w = tf.Variable(tf.random.truncated_normal(shape, stddev=0.1))
   
3.     if regularizer != None:
   
4.         tf.add_to_collection("losses", tf.keras.regularizers.l2(regularizer)(w))
   
5.     return w
   
6. 
   
7. def get_bias(shape):
   
8.     b = tf.Variable(tf.zeros(shape))
   
9.     return b

复制代码

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：

1. def test(mnist):
   
2.     x = tf.placeholder(tf.float32, shape=(None, mnist_forward.INPUT_NODE))
   
3.     y_ = tf.placeholder(tf.float32, shape=(None, mnist_forward.OUTPUT_NODE))
   
4.     y = mnist_forward.forward(x, None)
   
5. 
   
6.     ema = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
   
7.     ema_restore = ema.variables_to_restore()
   
8.     saver = tf.train.Saver(ema_restore)
   
9. 
   
10.     correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
    
11.     accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    
12. 
    
13.     while True:
    
14.         latest_checkpoint = tf.train.latest_checkpoint(mnist_backward.MODEL_SAVE_PATH)
    
15.         if latest_checkpoint:
    
16.             with tf.compat.v1.Session() as sess:
    
17.                 saver.restore(sess, latest_checkpoint)
    
18.                 global_step = int(latest_checkpoint.split("-")[-1])
    
19.                 accuracy_score = sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels})
    
20.                 print("After %d training step(s), test accuracy = %g" % (global_step, accuracy_score))
    
21.         else:
    
22.             print("No checkpoint file found")
    
23.             return
    
24. 
    
25. def main():
    
26.     mnist = tf.keras.datasets.mnist
    
27.     (x_train, y_train), (x_test, y_test) = mnist.load_data()
    
28.     x_test = x_test.reshape(-1, mnist_forward.INPUT_NODE) / 255.0
    
29.     y_test = tf.keras.utils.to_categorical(y_test, mnist_forward.OUTPUT_NODE)
    
30.     mnist.test = tf.data.Dataset.from_tensor_slices((x_test, y_test)).batch(10000)
    
31.     test(mnist)
    
32. 
    
33. if __name__ == "__main__":
    
34.     main()

复制代码

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：

1. def restore_model(testPicArr):
   
2.     with tf.Graph().as_default() as g:
   
3.         x = tf.placeholder(tf.float32, shape=(None, mnist_forward.INPUT_NODE))
   
4.         y = mnist_forward.forward(x, None)
   
5.         preValue = tf.argmax(y, 1)
   
6. 
   
7.         variable_averages = tf.train.ExponentialMovingAverage(mnist_backward.MOVING_AVERAGE_DECAY)
   
8.         variables_to_restore = variable_averages.variables_to_restore()
   
9.         saver = tf.train.Saver(variables_to_restore)
   
10. 
    
11.         with tf.compat.v1.Session(graph=g) as sess:
    
12.             latest_checkpoint = tf.train.latest_checkpoint(mnist_backward.MODEL_SAVE_PATH)
    
13.             if latest_checkpoint:
    
14.                 saver.restore(sess, latest_checkpoint)
    
15.                 preValue = sess.run(preValue, feed_dict={x: [testPicArr]})
    
16.                 return preValue[0]
    
17.             else:
    
18.                 print("No checkpoint file found in", mnist_backward.MODEL_SAVE_PATH)
    
19.                 return -1
    
20. 
    
21. def pre_pic(picName):
    
22.     img = Image.open(picName).convert('L')
    
23.     reIm = img.resize((28, 28), Image.LANCZOS)
    
24.     im_arr = np.array(reIm)
    
25.     threshold = 50
    
26.     im_arr = np.where(im_arr < threshold, 0, 255)
    
27.     nm_arr = im_arr.reshape(1, -1).astype(np.float32)
    
28.     img_ready = nm_arr / 255.0
    
29.     return img_ready
    
30. 
    
31. def application():
    
32.     testNum = int(input("input the number of test pictures:"))
    
33.     for i in range(testNum):
    
34.         testPic = input("the path of test picture:")
    
35.         testPicArr = pre_pic(testPic)
    
36.         preValue = restore_model(testPicArr)
    
37.         print("The prediction number is:", preValue)
    
38. 
    
39. def main():
    
40.     application()
    
41. 
    
42. if __name__ == "__main__":
    
43.     main()

复制代码

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。