luoyetx / JDA

C++ implementation of Joint Cascade Face Detection and Alignment.
BSD 3-Clause "New" or "Revised" License
184 stars 140 forks source link

How many landmarks does a feature need? #5

Closed GarrickLin closed 9 years ago

GarrickLin commented 9 years ago

在您的博客中有这样一句话

在 3000fps 中,由于随机森林是针对单个关键点的,所有随机树中使用到的特征点不会关联到其他关键点上,只在当前关键点的附近区域随机产生两个特征点,做像素差值来作为 Shape-index 特征。

另外在原文3000fps的section 3.3中提到

1) ...
2) we only consider the pixel features in the local region of a landmark 

上述都表明在文章3000fps中的做法是在训练一个landmark的cart的时候,只用了它本身附近的shape-index特征 但是在您的代码中,Feature是包含有两个landmark_id的,请问这样做的原因是什么呢?

GarrickLin commented 9 years ago

I'm sorry. I found the point in section 4.2 Joint Learning of Detection and Alignment of JDA.

luoyetx commented 9 years ago

3000fps and JDA use different methods of feature extraction, we can also try RCPR's way.