search

jessemelpolio / Faster_RCNN_for_DOTA

Code used for training Faster R-CNN on DOTA
https://arxiv.org/abs/1711.10398
Apache License 2.0
336 stars 102 forks source link

A question about the resnet_v1_101_rcnn_quadrangle #30

Closed qimw closed 5 years ago

qimw commented 5 years ago

In resnet_v1_101_rcnn_quadrangle.py, you define the bbox_weight as shape=(-1, 8 num_classes) at line 996 , and at line 1023 the output dim of fully connected layer is num_reg_classes 4. why there is a mismatch between those two dimension?

jessemelpolio commented 5 years ago

Oh, sorry for this misleading. The codes you mentioned at line 996 and line 1023 are never used under the end to end training scheme. get_symbol_rcnn and get_symbol_rpn function will in this case never be called. Only the codes in get_symbol function matter. I'll later fix the problem you mentioned. For now, you could just ignore it.

qimw commented 5 years ago

Oh, sorry for this misleading. The codes you mentioned at line 996 and line 1023 are never used under the end to end training scheme. get_symbol_rcnn and get_symbol_rpn function will in this case never be called. Only the codes in get_symbol function matter. I'll later fix the problem you mentioned. For now, you could just ignore it.

Thank you for your reply very much! :)