Retinaface-Cpp-mxnet

Update 2019.10.6

Faster-Retinaface-Cpp-mxnet:faster and lighter

Update 2019.9.9

Fixed bugs with demo.the demo is write in github file edit...so you know...

Update 2019.8.28

Upload Example....i forgot it at first...

Update

debug:flip result

debug:when use_lankmarks=false,output extract idx is wrong.

reduce packing and unboxing matrix times and time

No “vote” for the time being,i think it is not often to used,so i have no debug it,maybe.....Coming soon

Model from deepinsight/insightface/RetinaFace

deepinsight:

Pretrained Model: RetinaFace-R50 (baidu cloud or dropbox) is a medium size model with ResNet50 backbone. It can output face bounding boxes and five facial landmarks in a single forward pass.

WiderFace validation mAP: Easy 96.5, Medium 95.6, Hard 90.4.

To avoid the confliction with the WiderFace Challenge (ICCV 2019), we postpone the release time of our best model.

Third-party Models

yangfly:

RetinaFace-MobileNet0.25 (baidu cloud). WiderFace validation mAP: Hard 82.5. (model size: 1.68Mb)

Depend

opencv cpp lib

mxnet cpp lib

cuda 10.0

cudnn 7.5