How to get normal probabilities from predict_array on a single image?


#1

Is it possible to get normal probabilities (not log probabilities) on a single image prediction? The predict_array gives me logs probabilities where predict_batch gives me normal probabilities. How can I achieve such probabilities while making prediction on a single image?

Current code:
imt = val_tfms(im)
testmodel.precompute=False
log_preds = learn.predict_array(imt[None])
pred = np.argmax(preds)


#2

Just for others’ information, one could get the probabilities with the following formula:

probs = np.exp(log_preds)