In vggsound, we include the codes of Naive Fusion and our UMT in vggsound/av folder.

In UCF-101, we include the codes of Naive Fusion and our UMT in train_fusion.py and train_distillation.py, respectively.

About RGB-D segmentation, we implement UMT based on ESANet. And you can see our modification in RGB-D NYU-v2/rgbd_distill.py and some related python files.
