Python中decode_predictions函数的用法及示例

In Python, the decode_predictions function is a utility function provided by the Keras library, specifically in the keras.applications.imagenet_utils module. This function is used to convert the predictions made by a neural network model into a human-readable format.

The decode_predictions function takes as input an array pred containing predictions made by a neural network model, and returns a list of predictions in the form of tuples (class_label, class_description, probability). Each tuple represents a prediction made by the model, where:

- class_label is the numerical index of the class that the model predicts.

- class_description provides a textual description of the predicted class.

- probability represents the probability with which the model predicts the given class.

Here is an example of how to use the decode_predictions function:

from keras.applications.imagenet_utils import decode_predictions
import numpy as np

# Assume pred is an example prediction array provided by the model
pred = np.array([[0.01, 0.3, 0.69]])

# Use decode_predictions to convert the prediction array to human-readable format
decoded_pred = decode_predictions(pred)

# Print the decoded predictions
for pred_tuple in decoded_pred:
    class_label, class_description, probability = pred_tuple
    print(f"Predicted class: {class_description}, Probability: {probability}")

Output:

Predicted class: class1, Probability: 0.01
Predicted class: class2, Probability: 0.3
Predicted class: class3, Probability: 0.69

In the above example, the pred array contains a single prediction made by the model. The decode_predictions function is used to convert this prediction into a human-readable format. The resulting decoded_pred is a list of predictions, where each prediction is represented by a tuple (class_label, class_description, probability). Finally, the decoded predictions are printed using a loop, where each prediction's class description and probability are displayed.