zc2023 / TokenHPE

(CVPR 2023) TokenHPE: Learning Orientation Tokens for Efficient Head Pose Estimation via Transformers
MIT License
8 stars 0 forks source link

About Annotations #1

Open 7unahan opened 10 months ago

7unahan commented 10 months ago

Hello,

First of all, thank you for this excellent work. I am very confused about annotations. In your paper, Table 1, there is "free from extra annotations". What does it mean? There are .mat files in datasets but I understand that you do not use landmarks, could you help me for understand details because I want to work on Head Pose Estimation?

Best wishes

jkjgkg commented 3 months ago

为什么在BIWI训练集训练完了,在测试进行测试时候数据效果很差呢