Congratulations on your great work. However, I have a question why did you choose to apply delta for the missing modality in the middle of the network rather than at the input stage?
Thanks for your interest. During training the network, if we apply the delta at the input stage, it might disturb the training of the single-modal encoder.
Congratulations on your great work. However, I have a question why did you choose to apply delta for the missing modality in the middle of the network rather than at the input stage?