Lei Zhou
Results
1
issues of
Lei Zhou
We note that you use the features of last 4 layers from the encoder instead of intermediate layers (e.g. [5, 12, 18, 24] for vitl) as in some other works...