How to make a image pose estimation network

[thuanvh]

Hi all,
I am trying to use caffe to estimage the pose points of an image.
Input data is an image and label is a list of points (x,y) coordinate.
Could I model it by using caffe? and How do I ?

[thuanvh]

I found the answer in #512

[zirohut]

Does the label go like

.jpg x1_coordinate y1_coordinate x2_coordinate y2_coordinate

in the train.txt?

And may I ask how did you convert those data into hdf5 files? Will it work with lmdb?

Thanks !