Jia Guo
Jia Guo
@SthPhoenix Thanks! Did you make the feature maps shared? BTW, you can open a new repo to place this new model so that I can give a link to it,...
> > @SthPhoenix Thanks! Did you make the feature maps shared? > > No, shared feature maps seems to reduce accuracy more noticably. > > > BTW, you can open...
Shared feature map should be better by using GN, from my experiments of resnet based backbone.
I don't think the results are reliable because of 99.78% on lfw.
It's coming from another paper(git repo). Slight y-axis changes while doing left-right mirroring can bring some image-augmentation.
@AugustasMacys What we only need is the 5 landmarks from your input image, it's not about the original image size.
Can you please list your environment like OS/CUDA/MXNet versions?
Don't know much about windows. I can only suggest you try it on linux. And also fix the input size.
@k128 Try resizing/padding your input images to the same size.
@k128 MXNet inference engine may maintain a static networks pool, and create a new handler for each different input resolution in the meantime.