ScutterKey

here is some time cost using fcn and the model you provided. Are these right? time cost 0.744 pnet 0.733 rnet 0.009 onet 0.003 time cost 0.833 pnet 0.820 rnet...

@Seanlinx I use WIDER_train but gen_hard_examples.py is too slow in detect_face. Am i wrong

Could you release deploy.prototxt or train.prototxt? Thanks!

Maybe the problem auto_reset=True in speedometer

I find this source code in mxnet base_module.py, ``` for each epoch: if batch_end_callback is not None: batch_end_params = BatchEndParam(epoch=epoch, nbatch=nbatch, eval_metric=eval_metric, locals=locals()) for callback in _as_list(batch_end_callback): callback(batch_end_params) # ->...

@lsy17096535 你更新了prior_boxes_layer.cpp，能否把prototxt也更新一下。

I also get out of memory on V100 with batchsize64;

@jabooth Could you tell me the relationship between .jpg and .obj?

I use the follow code to eval celebA. However I get mAcc 0.8932 in celeba-val, mAcc 0.8856. It is not the same as paper claimed 0.9139 ``` import numpy as...

@seathiefwang Do you release your pretrained model?