search

wadimkehl / ssd-6d

Inference code and trained networks for SSD-6D
MIT License
310 stars 78 forks source link

Deep Learning of Local RGB-D Patches for 3D Object Detection and 6D Pose Estimation #2

Closed zgq91 closed 6 years ago

zgq91 commented 7 years ago

Hi: Recently looking at your "Deep Learning of Local RGB-D Patches for 3D Object Detection and 6D Pose Estimation" paper, where the network structure is not very clear. Is there any more detailed information?

wadimkehl commented 7 years ago

What exactly is not clear? There is a figure depicting the whole architecture and some text that explains every layer.

zgq91 commented 7 years ago

@wadimkehl input patch x(32x32x4): 4on behalf of what it means? rgb + depth this understanding is correct.What is the network output y structure?

wadimkehl commented 7 years ago

Since it is an autoencoder, input and output have the same shape.

zgq91 commented 7 years ago

@wadimkehl Can provide papers code, just look at the essay many places stumbled. If you can provide the code would be very much appreciated, my email zgqcontact@163.com.