laodar
laodar
It is just a regression task. Do you mean how to use mxnet to achieve this?
I will update the training results on a background agumented dataset. It may need several months.
@hyh21521038 using mtcnn(https://github.com/pangyupo/mxnet_mtcnn_face_detection) to get 5-point,then call the crop_and_predict(img,points)
> @mchong6 Have you solved this problem?
@nerdmiao are you using python3 ?
@nerdmiao Python 2.7.12
@ChaneWu Thanks, I have fixed it.
@canbaoburen I directly segment the input by separately reconstructing the two most likely digits, because if CapsNet works like what Hinton expects, the two longest DigitCaps Vectors will separately describe...
@af3113 Actually the two digit images segmented as showed in this README file are got by reconstructing from two longest DigitCap vectors. I think if the CapsNet really works as...