search

ijkguo / mx-rcnn

Parallel Faster R-CNN implementation with MXNet.
Other
669 stars 292 forks source link

AssertionError: Single batch only_ #55

Closed weiliuxm closed 7 years ago

weiliuxm commented 7 years ago

Hi, I employed train_end2end.py to train data. To train faster, I setconfig.TRAIN.BATCH_IMAGES = 2 in the file train_end2end.py. However, an error is reported.

_Traceback (most recent call last): File "train_end2end.py", line 186, in main() File "train_end2end.py", line 183, in main lr=args.lr, lr_step=args.lr_step) File "train_end2end.py", line 54, in train_net anchor_ratios=config.ANCHOR_RATIOS, aspect_grouping=config.TRAIN.ASPECT_GROUPING) File "/home/weiliu/projects/traffSgn/mx-rcnn/rcnn/core/loader.py", line 269, in init self.get_batch() File "/home/weiliu/projects/traffSgn/mx-rcnn/rcnn/core/loader.py", line 356, in get_batch data, label = get_rpn_batch(iroidb) File "/home/weiliu/projects/traffSgn/mx-rcnn/rcnn/io/rpn.py", line 46, in get_rpnbatch assert len(roidb) == 1, 'Single batch only' AssertionError: Single batch only

Any suggestion? Thank you in advance!

ijkguo commented 7 years ago

"Single batch only" --> "sorry, unimplemented"

breeze5428 commented 7 years ago

Thank you for reply.

Is config.TRAIN.BATCH_IMAGES not be able to set to be 2?

It works well when config.TRAIN.BATCH_IMAGES = 2in the file of train_end2end.py.

ijkguo commented 7 years ago

Only train_rcnn can do that.

breeze5428 commented 7 years ago

Thank you.