Steven Miao

[email protected]

Tencent Guangzhou, Guangdong, China. Working on computer vision, machine learning, deep learning, etc.

Results 3 comments of


                                            Steven Miao

I try to convert model from pytorch to caffe but a error occurred:

Yes, that is required.

About Train In ImageNet

Normally the initial loss of ImageNet should be around 6.9.

Do you provide pretrained weights of model_width=1.0 and model_width=2.0

My model didn't trained very well. I got a much worse accuracy compared with office paper. So I haven't published my own model.