Multitask cascaded architectures

facebookresearch / DensePose

A real-time approach for mapping all human pixels of 2D RGB images to a 3D surface-based model of the body

http://densepose.org

Other

6.97k stars 1.3k forks source link

Multitask cascaded architectures #62

Open Molength opened 6 years ago

Molength commented 6 years ago

How to achieve the task of Multi-task cascaded architectures in this paper 3.3?

zhiAung commented 6 years ago

I also have this question, I have not found anything related to multitasking cascading training in the training configuration file and code.

nvrv commented 6 years ago

@Molength, @zhiAung the cascaded architectures described in the paper are not a part of this code release.

Molength commented 6 years ago

Thanks for your reply ，I will try it myself.

troylhy1991 commented 5 years ago

Anybody has any idea of cross-cascading architecture as shown in figure 8 in the paper?

Q1: how can the two tensors (orange & green ones) be added up while the dimension (512x512x1x1 & 17x512x1x1) do not match? I assume 512 and 17 are # of output channels.

Q2: will the losses be calculated for both the refined and non-refined outputs?

Will appreciate some more details and explanation here :D