search

wei-mao-2019 / LearnTrajDep

code for learning trajectory dependencies for human motion prediction
MIT License
244 stars 45 forks source link

About Visualization of 3D data #8

Closed DanLuoNEU closed 4 years ago

DanLuoNEU commented 4 years ago

As it is mentioned in the paper, when you are using 3D data to train the network, you eliminate the Global Rotation and Translation, which are the first 6 dimensions of the exponential map. 

# remove global rotation and translation
the_seq[:, 0:6] = 0
p3d = expmap2xyz_torch(the_seq)

So when I am using your training or testing 3D data to do the visualization, the whole skeleton is a sloping one. Like the figure below shows: image But as you show in the paper, the skeleton is totally fine, would you help me with some tools or codes that you use to generate a normal skeleton from your training or test 3D data?

wei-mao-2019 commented 4 years ago

Hi @DanLuoNEU,

I think you may need to adjust the view point if you are using matplotlib.pyplot. You can check the official document for that.

Cheers Wei