Combining the output of 3 Bayesian Convolutional NNs

vision
2

Double post from here with follow-up.
Let’s please keep the discussion in one place (here) to avoid re-debugging the same issue.