bharatsingh430 / deformable-convnets Goto Github PK

@bharatsingh430

How to use the soft-mns in training or testing?

@bharatsingh430 I started to train Faster RCNN on my custom dataset, and it has passed 50 epoches, but when I want to test the saved model then I can't find it, so the question comes out where the models saved?

HI~
I am a fresher in mxnet.
It is very appreciate that you release this work! Is it possible that you add README about how to train the model more detail?
Thank you in advance!

Is any pretrained model available for PASCAL VOC dataset?
Thanks!

Hi @bharatsingh430 ,

does your code support training for Deformable ConvNets?

Thanks.

how to do the multi scale testing?

@bharatsingh430 When run ./rfcn/demo.py, in order to load your trained params: rfcn_dcn_coco-0008.params, I change the line 95 from:
arg_params, aux_params = load_param(cur_path + '/../model/' + ('rfcn_dcn_coco' if not args.rfcn_only else 'rfcn_coco'), 0, process=True)
to:
arg_params, aux_params = load_param(cur_path + '/../model/' + ('rfcn_dcn_coco' if not args.rfcn_only else 'rfcn_coco'), 8, process=True)
While this is an error: Check failed: from.shape() == to->shape() operands shape mismatchfrom.shape = (60,) to.shape=(48,)
Is it due to the parameters of anchor? How should I change the config file.
Thanks.

Could u tell me which version of MXnet and CUDA do u use? Thank u~~