add IMAGE_CHANNEL_COUNT class variable to config to make it easier to use Mask_RCNN for non 3-channel images

Fixed comment on GPU_COUNT

Fix bugs of not showing boxes in 1.b RPN Predictions.
TF 1.9 introduces "ROI/rpn_non_max_suppression/NonMaxSuppressionV3:0", so the original code can't work.

Unified regex expression for both, Linux and Windows.

- Current implementation creates self.log_dir in set_log_dir() function,
  which creates an empty log directory if none exists. This causes
  find_last() to fail after creating a model because it finds this new
  empty directory instead of the previous training directory.
- New implementation moves log_dir creation to the train() function to
  ensure it is only created when it will be used.

VIA has changed JSON formatting in later versions. Now instead of a dictionary, "regions" has a list, see the issue #928

Add the possibility to exclude some sources from augmentation by passing a list of sources. This is useful when you want to retrain a model having few images.

Add an optional parameter for calling a list of keras.callbacks to be add to the original list.

Improve previous commit to not break on older versions of Keras.

Raise clear error if last training weights are not foundIf using the --weights=last (or --model=last) to resume trainingbut the weights are not found now it raises a clear error message.

modified comment for image augmentation line import to include correct 'pip3 install imgaug' instructions

This allows one to set a callable in Config.BACKBONE to use a custom
backbone model.

Fix matterport/Mask_RCNN#170

Print a message when re-starting from saved epoch

fix typo

Reported here: https://github.com/matterport/Mask_RCNN/pull/500