我有CNN模型,它有4个输出节点,我试图计算混淆矩阵,这样我就可以知道各个类的准确性.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”)给出了一个错误,错误是TypeError:attr’T’的DataType float32不在允许值列表中:int32,int64
我已经尝试过将类型转换为int32内部函数提到的def评估(logits,labels),它在计算correct_prediction =时给出了另一个错误:…作为TypeError:’InTopK’Op的输入’预测’的类型为int32,与预期类型不匹配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})
这是我的代码片段
解决方法:
你可以简单地使用Tensorflow的confusion matrix.我假设y是你的预测,你可能有也可能没有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)
如果你打印(混乱),你得到
[[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(混淆)没有打印混淆矩阵,则使用print(confusion.eval(session = sess)).这里是sess是TensorFlow会话的名称.
标签:python,tensorflow,confusion-matrix