bug fixes to the rpn - be careful to keep the zeroth axis for the batch

broadinstitute / keras-rcnn

Keras package for region-based convolutional neural networks (RCNNs)

Other

553 stars 223 forks source link

Closed yhenon closed 6 years ago

yhenon commented 6 years ago

Some fixes for the RPN.

It's important to keep the zeroth axis for the number of samples in the batch, or keras will have many issues. This was not the case in the current code.
Updated _anchor_target to use metadata as the input instead of using int_shape on the image. Should make things easier.

0x00b1 commented 6 years ago

Awesome! Thanks, @yhenon.