baseline result

[Jiawei0o0]

Hi， thanks for sharing your wonderful work.
I got a wired result with your code when only the T1ce modality was used. The input shape was set as 128128128 in my experiment. Can you share your baseline result with the T1ce modality?
By the way, you have fixed the input shape in the data.py(line 132) which makes the (input shape: [160, 192, 128] in the config.yml ) useless.

[scut-cszcl]

Hello, whether this work train a dedicated model for all possible subsets of available modalities? If not, how does it handle the missing modal input, is it filled with zero?

[Jiawei0o0]

You are right. It trains a model separately for different situations.