search

lorenmt / mtan

The implementation of "End-to-End Multi-Task Learning with Attention" [CVPR 2019].
https://shikun.io/projects/multi-task-attention-network
MIT License
665 stars 108 forks source link

Question about the architecture #27

Closed minygd closed 4 years ago

minygd commented 4 years ago

Hello, I have a question about the architecture of the model. In Fig.2, is the pool in Attention Module for Encoder the same as the same stage in VGG? And so as the samp in Decoder part. Thanks for your time!

lorenmt commented 4 years ago

Hi,

I am not sure I understand your question correctly. Do you want to ask whether the design of the pooling and sampling function in the attention module and the shared network is the same? If so, yes they are the same. pool uses MaxPool(2, 2) and samp uses Bilinear sampling.

For detailed design, I suggest checking out the original code.

minygd commented 4 years ago

OK, thanks. I got it!