nateraw commited on
Commit
48019b4
1 Parent(s): 2113bc2

Update dino/app.py

Browse files
Files changed (1) hide show
  1. dino/app.py +2 -2
dino/app.py CHANGED
@@ -51,10 +51,10 @@ def func(resize, video):
51
  return f'{output_dir}/stacked.mp4'
52
 
53
  title = "Interactive demo: DINO"
54
- description = "Demo for Facebook AI's DINO, a new method for self-supervised training of Vision Transformers. Using this method, they are capable of segmenting objects within an image without having ever been trained to do so. This can be observed by displaying the self-attention of the heads from the last layer for the [CLS] token query. This demo uses a ViT-S/8 trained with DINO. To use it, simply upload an image or use the example image below. Results will show up in a few seconds."
55
  article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2104.14294'>Emerging Properties in Self-Supervised Vision Transformers</a> | <a href='https://github.com/facebookresearch/dino'>Github Repo</a></p>"
56
  iface = gr.Interface(fn=func,
57
- inputs=[gr.inputs.Slider(120, 420, 20, label="resize"), gr.inputs.Video(label="input video")],
58
  outputs='video',
59
  title=title,
60
  description=description,
 
51
  return f'{output_dir}/stacked.mp4'
52
 
53
  title = "Interactive demo: DINO"
54
+ description = "Demo for Facebook AI's DINO, a new method for self-supervised training of Vision Transformers. Using this method, they are capable of segmenting objects within an image without having ever been trained to do so. This can be observed by displaying the self-attention of the heads from the last layer for the [CLS] token query. This demo uses a ViT-S/8 trained with DINO. To use it, upload a video file. Right now we only suggest using .mp4 files."
55
  article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2104.14294'>Emerging Properties in Self-Supervised Vision Transformers</a> | <a href='https://github.com/facebookresearch/dino'>Github Repo</a></p>"
56
  iface = gr.Interface(fn=func,
57
+ inputs=[gr.inputs.Slider(120, 480, 20, label="resize"), gr.inputs.Video(label="input video")],
58
  outputs='video',
59
  title=title,
60
  description=description,