Question

我有CNN模型，它有4个输出节点，我正在尝试计算混淆矩阵，以便我可以知道单个类的准确性。我能够计算出整体的准确性。在链接here中，Igor Valantic给出了一个可以计算混淆矩阵变量的函数。它在correct_prediction = tf.nn.in_top_k(logits, labels, 1, name="correct_answers")给出了错误，错误为TypeError: DataType float32 for attr 'T' not in list of allowed values: int32, int64

我已经尝试对提到def evaluation(logits, labels)的函数内部的int32进行类型转换，它在计算correct_prediction = ...时给出了另一个错误TypeError:Input 'predictions' of 'InTopK' Op has type int32 that does not match expected type of float32

如何计算这种混淆矩阵？

sess = tf.Session()
model = dimensions() # CNN input weights are calculated 
data_train, data_test, label_train, label_test =  load_data(files_test2,folder)
data_train, data_test, = reshapedata(data_train, data_test, model)
# input output placeholders
x  = tf.placeholder(tf.float32, [model.BATCH_SIZE, model.input_width,model.input_height,model.input_depth]) # last column = 1 
y_ = tf.placeholder(tf.float32, [model.BATCH_SIZE, model.No_Classes])
p_keep_conv = tf.placeholder("float")
# 
y  = mycnn(x,model, p_keep_conv)
# loss
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(y, y_))
# train step
train_step = tf.train.AdamOptimizer(1e-3).minimize(cost)
correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
true_positives, false_positives, true_negatives, false_negatives = evaluation(y,y_)
lossfun = np.zeros(STEPS)
sess.run(tf.global_variables_initializer())

for i in range(STEPS):
    image_batch, label_batch = batchdata(data_train, label_train, model.BATCH_SIZE)
    epoch_loss = 0
    for j in range(model.BATCH_SIZE):
        sess.run(train_step, feed_dict={x: image_batch, y_: label_batch, p_keep_conv:1.0})
        c = sess.run( cost, feed_dict={x: image_batch, y_: label_batch, p_keep_conv: 1.0})
        epoch_loss += c
    lossfun[i] = epoch_loss
    print('Epoch',i,'completed out of',STEPS,'loss:',epoch_loss )
 TP,FP,TN,FN = sess.run([true_positives, false_positives, true_negatives,  false_negatives], feed_dict={x: image_batch, y_: label_batch, p_keep_conv:1.0})

这是我的代码段

Answer 1

您只需使用Tensorflow的confusion matrix即可。我认为y是您的预测，您可能有也可能没有num_classes（这是可选的）

y_ = placeholder_for_labels # for eg: [1, 2, 4]
y = mycnn(...) # for eg: [2, 2, 4]

confusion = tf.confusion_matrix(labels=y_, predictions=y, num_classes=num_classes)

如果你print(confusion)，你会得到

  [[0 0 0 0 0]
   [0 0 1 0 0]
   [0 0 1 0 0]
   [0 0 0 0 0]
   [0 0 0 0 1]]

如果print(confusion)未打印混淆矩阵，请使用print(confusion.eval(session=sess))。这里sess是您的TensorFlow会话的名称。

Answer 2

import tensorflow as tf     
y = [1, 2, 4]
y_ = [2, 2, 4]

con = tf.confusion_matrix(labels=y_, predictions=y )
sess = tf.Session()
with sess.as_default():
        print(sess.run(con))

输出结果为：

[[0 0 0 0 0]
[0 0 0 0 0]
[0 1 1 0 0]
[0 0 0 0 0]
[0 0 0 0 1]]

如何在张量流中创建混淆矩阵进行分类

2 个答案: