Mastering Computer Vision with TensorFlow 2.x

Decoding predictions

The ImageNet data has 1,000 different classes. A neural network such as Inception trained on ImageNet will output the class as an integer. We need to convert the integer to a corresponding class name using decoding. For example, if the integer value output is 311, we need to decode what 311 means. With decoding, we will know that 311 corresponds to a folding chair.

The code for decoding predictions is as follows:

from tensorflow.keras.applications.vgg16 import decode_predictionsfrom tensorflow.keras.applications.resnet50 import decode_predictionsfrom tensorflow.keras.applications.inception_v3 import decode_predictions

The preceding code maps the class integers to class names using the decode_predictions command. Without ...