Codes of Facilitating Multimodal Classification via Dynamically Learning Modality Gap

Here is the official PyTorch implementation of ''Facilitating Multimodal Classification via Dynamically Learning Modality Gap''

Paper Title: "Facilitating Multimodal Classification via Dynamically Learning Modality Gap"

For replication inquiries or issues, feel free to contact us via email at [[email protected]] or [[email protected]].

Code instruction

Data Preparation

The original datasets can be found: CREMA-D, Kinetics-Sounds,

Run

The code use CREMA-D dataset as example. You can simply run the code using:

python Crema_epoch.py