Hugging Face's logo

fiskenai
/
Geo-Sign

Video-Text-to-Text
Transformers
Chinese
sign-language-translation
skeleton-based
hyperbolic-geometry
mT5
Model card Files
xet
Community
1
fiskenai commited on
Commit
3059c07
·
verified ·
1 Parent(s): 9d676ab

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -1
README.md CHANGED
@@ -28,13 +28,15 @@ Edward Fish, Richard Bowden, CVSSP – University of Surrey (arXiv:2506.00129, M
28
  ## Code Use
29
  Download the weights and data labels from the files section of this repo and add them to the github repository <https://github.com/ed-fish/geo-sign>.
30
 
 
 
31
  `Data -> ./Data`
32
 
33
  `best.pth -> ./checkpoints/best.pth`
34
 
35
  `pretraining.pth -> ./checkpoints/pretraining.pth`
36
 
37
- `pretrained_weight/ -> ./pretrained_weight`
38
 
39
  ## TL;DR
40
  Geo-Sign projects pose-based sign-language features into a learnable **Poincaré ball** and aligns them with text embeddings via a geometric contrastive loss.
 
28
  ## Code Use
29
  Download the weights and data labels from the files section of this repo and add them to the github repository <https://github.com/ed-fish/geo-sign>.
30
 
31
+ You will also need the base-mt5 model from <https://huggingface.co/google/mt5-base> and put it in the pretrained_weight folder.
32
+
33
  `Data -> ./Data`
34
 
35
  `best.pth -> ./checkpoints/best.pth`
36
 
37
  `pretraining.pth -> ./checkpoints/pretraining.pth`
38
 
39
+ `<https://huggingface.co/google/mt5-base> -> ./pretrained_weight`
40
 
41
  ## TL;DR
42
  Geo-Sign projects pose-based sign-language features into a learnable **Poincaré ball** and aligns them with text embeddings via a geometric contrastive loss.