DONG
DONG
In [link](https://github.com/horovod/horovod/blob/master/examples/pytorch/pytorch_mnist.py), adasum optimizer do not support method skip_synchronize and synchronize when train with mixed precision. without that methods loss will all become nan.
Thank you for he sphereface code! but I don't know how to get to get the coord5point in face_align_demo.m .
Thank for your work! I want to convert arcface to tensorrt, but [nn.BatchNorm1d(feature_dim))](https://github.com/wujiyang/Face_Pytorch/blob/master/backbone/arcfacenet.py#L146) do not support 2D in tensorrt. the error message is " what(): No converter registered for op...
Thank you for the paper and the source code! when trained with noise_tolerant+(arcface-loss)(margin = 0.5) , the loss and accuracy(nearly 0) did not converge ? However, when the margin was...
Thank your for the good work! I use the docker file to train "CondInst_MS_R_50_1x" and change the wight from "detectron2://ImageNetPretrained/MSRA/R-50.pkl" to local R-50.pkl(downloaded from https://dl.fbaipublicfiles.com/detectron2/ImageNetPretrained/MSRA/R-50.pkl), the performance is as below...
Hello，I can not get the BIWI dataset after send the License agreement email, Can anyone share the BIWI dataset? Thank you very much!
thank your for your great work! I have benefited much from Parsing-rcnn. I have been looking forward to RP-R-CNN for a long time, this paper still give me much surprise....
The number of data labels is 4. the eval loss increase and the train loss decrease. I think it is overfit. Code is as below: # code ```from datasets import...