Closed GMahdi closed 4 years ago
The idea is get score for all 20 objects not the only one which have the highest score; my understanding at some place it is being done argmax to get the label but how to get all the list.
The predictions are not filtered by argmax but filtered out by threshold as: confidence < bbox_threshold
.
You can set bbox_threshold to a low value (e.g. 0.05) to get all of the objects.
https://github.com/kentaroy47/frcnn-from-scratch-with-keras/blob/master/test_voc.py#L177
Thanks for the great work; I am trying to get distributions for all objects of test dataset of Pascal 07. It means instead of giving maximum score for a bounding box, I want to print score for scores for all 20 objects. Is there a way to get it? thanks