如何在张量流中给定的CNN中找到预测概率？

Question

I am very new to TensorFlow. Let us assume that I already have a trained convolutional neural network, now I give one new data to this CNN, and I want to see what's the prediction probability in each class. (eg, the CNN is for handwriting 0-2, now I give a new data 2 to this trained CNN, the prediction probability should give me something like 0.01 for class 0, 0.02 for class 1, and 0.97 for class 2)

May I ask someone advise me, what's the right code to do that in TensorFlow (1.13.1) for python? Sorry about the elementary level question.

I am using the online example MNITS code,

import numpy as np
import tensorflow as tf


def cnn_model_fn(features, labels, mode):
  input_layer = tf.reshape(features["x"], [-1, 28, 28, 1])
  conv1 = tf.layers.conv2d(inputs=input_layer, filters=30, kernel_size=[5, 5], padding="same", activation=tf.nn.relu)
  pool1 = tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
  pool2_flat = tf.reshape(pool1, [-1, 14 * 14 * 30])
  dense = tf.layers.dense(inputs=pool2_flat, units=1000, activation=tf.nn.relu)

  dropout = tf.layers.dropout(inputs=dense, rate=0.4, training=mode == tf.estimator.ModeKeys.TRAIN)

  logits = tf.layers.dense(inputs=dropout, units=10)

  predictions = {
      # Generate predictions (for PREDICT and EVAL mode)
      "classes": tf.argmax(input=logits, axis=1),
      # Add `softmax_tensor` to the graph. It is used for PREDICT and by the
      # `logging_hook`.
      "probabilities": tf.nn.softmax(logits, name="softmax_tensor")
  }

  if mode == tf.estimator.ModeKeys.PREDICT:
    return tf.estimator.EstimatorSpec(mode=mode, predictions=predictions)

  # Calculate Loss (for both TRAIN and EVAL modes)
  loss = tf.losses.sparse_softmax_cross_entropy(labels=labels, logits=logits)

  # Configure the Training Op (for TRAIN mode)
  if mode == tf.estimator.ModeKeys.TRAIN:
    optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
    train_op = optimizer.minimize(
        loss=loss,
        global_step=tf.train.get_global_step())
    return tf.estimator.EstimatorSpec(mode=mode, loss=loss, train_op=train_op)

  # Add evaluation metrics (for EVAL mode)
  eval_metric_ops = {
      "accuracy after all": tf.metrics.accuracy(
          labels=labels, predictions=predictions["classes"])}
  return tf.estimator.EstimatorSpec(
      mode=mode, loss=loss, eval_metric_ops=eval_metric_ops)


def main(unused_argv):
  model_path = "/tmp/mnist_convnet_model"

  # Load training and eval data
  mnist = tf.contrib.learn.datasets.load_dataset("mnist")
  train_data = mnist.train.images  # Returns np.array
  train_labels = np.asarray(mnist.train.labels, dtype=np.int32)
  eval_data = mnist.test.images  # Returns np.array
  eval_labels = np.asarray(mnist.test.labels, dtype=np.int32)

  # Create the Estimator
  mnist_classifier = tf.estimator.Estimator(
      model_fn=cnn_model_fn, model_dir=model_path)

  # Set up logging for predictions
  # Log the values in the "Softmax" tensor with label "probabilities"
  logging_hook = tf.train.LoggingTensorHook(
      tensors={"probabilities": "softmax_tensor"}, every_n_iter=50)

  # Train the model
  train_input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
      x={"x": train_data},
      y=train_labels,
      batch_size=100,
      num_epochs=None,
      shuffle=True)
  mnist_classifier.train(
      input_fn=train_input_fn,
      steps=5000,
      hooks=[logging_hook])

  # Evaluate the model and print results
  eval_input_fn = tf.compat.v1.estimator.inputs.numpy_input_fn(
      x={"x": eval_data}, y=eval_labels, num_epochs=1, shuffle=False)
  eval_results = mnist_classifier.evaluate(input_fn=eval_input_fn)
  print(eval_results)


if __name__ == "__main__":
  tf.app.run()

Answer 1

Call the predict method of your estimator (ie, mnist_classifier ) and set predict_keys="probabilities" .

The predict method only runs the inference (unlike evaluate without evaluation). Setting the key will choose the correct tensor from the dictionary called predictions that you have in the cnn_model_fn method.