DLIB：针对 194 个地标训练 Shape_predictor（helen 数据集）

我正在训练DLIB's 形状预测器对于 194 个面部特征点，使用海伦数据集用于通过以下方式检测人脸特征点face_landmark_detection_ex.cppdlib 库。

现在它给了我一个sp.dat大约的二进制文件45 MB与给定的文件相比较少（http://sourceforge.net/projects/dclib/files/dlib/v18.10/shape_predictor_68_face_landmarks.dat.bz2）用于 68 个面部特征点。在培训中

• 平均训练误差：0.0203811
• 平均测试误差：0.0204511

当我使用训练数据来获取面部标志位置时，我得到的结果是..

与 68 个地标的结果偏差很大

68个地标图像：

Why?

好吧，看来你还没有读过code评论 （？）：

shape_predictor_trainer trainer;
// This algorithm has a bunch of parameters you can mess with.  The
// documentation for the shape_predictor_trainer explains all of them.
// You should also read Kazemi's paper which explains all the parameters
// in great detail.  However, here I'm just setting three of them
// differently than their default values.  I'm doing this because we
// have a very small dataset.  In particular, setting the oversampling
// to a high amount (300) effectively boosts the training set size, so
// that helps this example.
trainer.set_oversampling_amount(300);
// I'm also reducing the capacity of the model by explicitly increasing
// the regularization (making nu smaller) and by using trees with
// smaller depths.  
trainer.set_nu(0.05);
trainer.set_tree_depth(2);

看看卡泽米纸，ctrl-f 字符串“参数”并读取...