Can

Same question，one polygon seems unable to cover these situations:  

How to get 3d landmarks indies with respect to 2D landmark (106 landmarks, or 1000 predefined 2D landmarks) ?

这类FOMM方法好像不太好解决这个问题，要么像作者提到的：1、做人脸融合，2、局部区域生成（比如只生成嘴巴区域，然后后处理融合回原图），3、或者扩大生成区域，不融合回原图。难，我也遇到了：https://github.com/NetEase-GameAI/Face2FaceRHO/issues/25

https://github.com/daikankan/MTCNN_face_detection_alignment

dynamic_lmk_faces_idx is selected by pose as follows, right ? 

Hi, @cdicle I'm also trying the training process of mtcnn, I guess there are many details and tricks that I don't know, expecially for landmark localization regression in multi-task training,...

any update? I use pytorch_1.13.1, convert pytorch model to onnx use torch.onnx.export(opset_version=16), then convert onnx model to tensorrt-8.6.1.6, and get error: Error[3]: [network.cpp::addGridSample::1537] Error Code 3: API Usage Error (Parameter...

Me too, my question is how to implement nms when batch_size > 1 using tensorflow, tf.image.non_max_suppression seems can't.

@agupta54 just opencv: circle rectangle and puttext