Jia Guo comments

Results 187 comments of


                                            Jia Guo

scrfd.py onnx inference doesn't use GPU

remove both of them, then re-install onnxruntime-gpu

jmlr training code vs description in arxiv paper

Hi, 1. The training input size is 256, please see https://github.com/deepinsight/insightface/blob/master/reconstruction/jmlr/configs/s1.py#L18 2. you are right, we will fix the code soon.

jmlr training code vs description in arxiv paper

The number 512 is just the size of the image in rec file, not the training size.

problem about the preprocess of face image before feed it into face recognition onnx model

https://github.com/deepinsight/insightface/blob/master/python-package/insightface/utils/face_align.py#L70

Error in train phase (libnccl.so.2: cannot open shared object file: no such file or directory)

You have to install cuDNN. By the way, I recommend you to use the PyTorch version of arcface training code.

scrfd的检测的人脸置信度低

正常分数绝对值没太大意义

Incorrect fc weights in [Resnet34, Arcface, MS1MV3] pretrained models

the softmax fc weights should not be saved in our pretrained models.

Where can i find generate_edge.py?

@QingpingZheng

MFR Ongoing测试服务器挂机的问题

升级cudnn后，cuda10.2+onnxruntime-1.6-gpu可以正常工作吗？

MFR Ongoing测试服务器挂机的问题

不是cuda的适配性，而是onnxruntime-1.6-gpu可否正常执行。1.6-gpu默认要求的环境是cudnn 8.0.5