this is a VGG3D algorithm used for RGB-D object recognition, but the version is old. There are many difference for the new julia version. I think that will help julia with some improvement also.
https://github.com/samanzia/rgbd-convnet
the following is the pappers.
and I’m debugging it for a long time. if you can solve it, i will be appreciate
May I ask what your question is exactly?
@pkofod I’d guess related to I need some help · Issue #1 · samanzia/rgbd-convnet · GitHub
Have you checked out KNet - beginning deep learning with 100 lines of julia.pdf
It would probably be easier to get the VGG3D running on the CPU first
sorry, it’s useful for me at first. And it may have some help to julia. thanks your suggestions.