Solution for How to get percentage prediction for each class from TensorFlow?
is Given Below:
I’m creating a neural network for image classification and i have the following classes:
[class1, class2, class3]
I have a model based on VGG with the last layer of the model being modified as follows:
predictions = layers.Dense(int(len(['class1', 'class2', 'class3'])), activation='softmax')(x)
Then I compile the model as follows:
model.compile(loss="categorical_crossentropy", optimizer=Adam(), metrics=['accuracy'])
Then I predict the class according to the input image:
img = load_img(input_path, target_size=(224, 224)) img = np.asarray(img) / 255. img = cv2.resize(img, (224, 224)) img = img.reshape(1, 224, 224, 3) prediction = model.predict(img)
However, it does come back to me as a kind of prediction:
[[5.7733614e-29 2.5203591e-28 3.1751932e-38]]
I would like to understand what this data means and how it can be interpreted as a percentage, so that I can say that class1 is 59%, class2 is 20% and class3 is 21%.
I mean, I’d like the output to be like this:
[[0.59 0.20 0.21]]
Am I doing something wrong, or am I just working with the data incorrectly?
Thank you very much.
Edit: For Imagenet only.
You should use
decode_predictions on top of
predict to get more human-readable results.
To illustrate, the output may look like after calling
3.60479535e-05 2.34715412e-06 9.06522095e-04 1.34255132e-03
[[('n01944390', 'snail', 0.24689779), ('n01943899', 'conch', 0.2438357), ('n03944341', 'pinwhe ...
Please try the below snippet.
from tensorflow.keras.applications.vgg16 import decode_predictions prediction = model.predict(img) # You can use the one below. # print( 'Predicted:', decode_predictions( prediction, top=5 ) ) # or... results = decode_predictions(prediction) for result in results: print( result ) # prints the accuracy levels of each class