How To Get The Class With Highest Confidence From A Tensor?

I want to obtain the class with the highest confidence. Heres the code that performs classification task: names = ['class A', 'class B', 'class C'] def classify_face(image): d

Solution 1:

It seems you will always be looking at the tensor pred[0]. So let pred be:

pred = torch.tensor([[212.38568, 117.47020, 339.35773, 266.00513, 0.74144, 2.00000],
                     [214.60651, 118.50694, 339.90192, 265.91696, 0.94277, 0.00000]])

The index of the prediction with highest confidence is:

i = torch.argmax(pred[:, 4])

Therefore, you just have to get the last value at that index:

pred[i, -1]

And the class name will be names[int(pred[i, -1])].