Multilabel prediction probabilities

Part 1 (2020)
1

Is it possible to get multilabel probabilities when we are using model on a test set (something not seen before)? Like for example, image A will return 0.3 probability of tree, 0.2 probability of a flower and .15 probability of a park bench (the image contains a tree, a flower and a bench). Image B will return 0.15 probability of a lake and and 0.5 probability of a swan (the image has a lake and a swan on it). If it is possible to do, how can we do it in practical terms with get prediction?

2

Any answer for this?

3

I discuss what’s needed for such an approach here:

1 Like