关于mxnet-face about mxnet-face HOT 8 CLOSED

hello,i only changed the lr and batchsize from its orinigal paper,because i use multigpu for training. so whats failure do you mean in your training?

from mxnet-face.

i only add stop_factor_lr when use factor lr,please look #1

from mxnet-face.

你怎么把MXNET用的这么6?

from mxnet-face.

because MXNet is a cool tool for dl.:)

from mxnet-face.

你是按照设什么标准进行人脸的矫正的呢？

from mxnet-face.

@ZHAIXINGZHAIYUE in the training phase, please refer to lfw.py, and in the testing phase, first use lfw.py to detect and align, then correct some error detection by hand(this is usually occur when two man in the lfw image, but the detection get the wrong face ).

from mxnet-face.

请教你个问题，为什么要把144＊144大小的图像随机crop到128*128？随机crop的作用和好处到底是什么呢？谢谢。

from mxnet-face.

data augment

from mxnet-face.