Open soham24 opened 4 years ago
I added support for the blazeface architecture. However you'll have to convert a model yourself, I only added the ability to do channel padding which is used by blazeface and what was missing in NCNN.
@soham24 I wonder if it worked? Thanks!
@Timen thanks for the update. @yeswecan Sorry, I have not tried the conversion yet
https://github.com/hollance/BlazeFace-PyTorch I am getting nan at output After converting via onnx simplifier
Has anyone tried to port this to ncnn