-
Author, thank you for your work. Here I would like to ask, is there any difference between the pre-training model you used in the stage of image feature extraction (**swin_large_patch4_window12_384_22…
-
In this repository, patch merging is implemented with nn.Unfold, but it is expected to behave differently than the official implementation.
https://github.com/microsoft/Swin-Transformer/blob/6ded25…
-
Hello -- really appreciate your work! I was able to get a ResNet 50 model to train perfectly well on my custom dataset using your config files and confirmed that Stable DINO / R50 is better than DINO …
-
Hi Hu Cao,
First, thanks for your wonderful work.
I'm wondering that how to revise your code to adapt to the 384x384 input image. The original Swin Transformer provides 384x384 Swin-B weight, …
-
After I loaded the pretrained model file (namely 'upernet_swin_small_patch4_window7_512x512.pth'), an error was raised as the follow. What is different between my custom dataset with ADE20K is that th…
-
Dear reaserchers, please also consider checking our newly introduced face inpainting method to address the symmetry problems of general inpainting methods by using swin transformer and semantic aware …
-
Hi, Thanks for your nice work! @kevinlin311tw
However, could you please report the performance of adopting the Video Swin Transformer as an offline-extracted video encoder?
In Table 2, othe…
-
Greetings!
I read your paper with great interest and am trying to reproduce some of your experiments.
I want to reproduce your vanilla KD setting using strategy B1, B2, B3 based on your DIST_KD…
-
The demo produced 2 errors when running.
`FileNotFoundError: [Errno 2] No such file or directory: './model/MEFL.py'
`
MEFL.py needs to be copied from the main repo.
Then once this is fixed:
`…
-
When trying to load the checkpoint after SLL pretraining (1 epoch to test) with MoBY I get this error after using the `--pretrained` flag pointing to the checkpoint (I've tried ckpt_epoch_0.pth and ch…