Why use 3D Conv/Deconv for gate Conv?

tsunghan-wu / Depth-Completion

🍀 Official pytorch implementation of "Indoor Depth Completion with Boundary Consistency and Self-Attention. Huang et al. RLQ@ICCV 2019."

https://arxiv.org/abs/1908.08344

MIT License

181 stars 18 forks source link

Why use 3D Conv/Deconv for gate Conv? #6

Closed mzy97 closed 3 years ago

mzy97 commented 4 years ago

Are you using a temporal sequence as the input of the network?

tsunghan-wu commented 4 years ago

NO, we didn't use temporal sequence as the input of the network. In our code, we passed conv_by='2d' to class GatedConv (see code here). Thus, the convolution operation is done on 2D space. If you have any questions, feel free to ask us for more detailed.

Best, Tsung-Han Wu

tsunghan-wu commented 3 years ago

Close for long inactivity