training data selection

[vijosam]

In the function get_top_confidences in matcher.py, you choose the top_k confidences bounding boxes including background, I think it is not a good strategy, it may lead to train negative examples that have already been trained. I think the top_k foreground bounding boxes should be chosen, am I right?

[hjptriplebee]

You are right.
The condition "np.argmax(predLabels[i]) != classNum" has ruled out the possibility of choosing background as negative examples, but if the top k examples are almost background, the number of negative examples will be reduced greatly.