about the meshgraphormer architecture problem

[zhangzhenyu-zzy]

Hello, I want to know two things,

1. I want to prove whether the input in the meshgraphormer framework and the information in the attention mechanism are as I drew the picture
   

2. I want to know that grid features are obtained by extracting features through CNN. Its dimension is 49×1024. It is not marked in the paper that it uses upsampling. How did it transform from 49×1024 to 49×2051?
   

[kevinlin311tw]

Q1: There are (49+14+431) input tokens. All tokens have 2051-Dim.

Q2: We apply MLP for the transform. In Sec 4.1, we mentioned that we apply MLP to make the size of all input tokens consistent. After that, all input tokens have 2051-Dim.

[zhangzhenyu-zzy]

thank you very much！