adirik commited on
Commit
91391e7
1 Parent(s): b357922
Files changed (1) hide show
  1. app.py +2 -3
app.py CHANGED
@@ -29,10 +29,9 @@ description = """
29
  </div>
30
  <div class="text">
31
  <p>Gradio demo for <a href="https://huggingface.co/docs/transformers/main/en/model_doc/align">ALIGN</a>,
32
- as introduced in <a href="https://arxiv.org/abs/2102.05918"></a><i>"Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision
33
- "</i>. ALIGN features a dual-encoder architecture with EfficientNet and BERT as its text and vision encoders, and learns to align visual and text representations with contrastive learning.
34
  Unlike previous work, ALIGN leverages a massive noisy dataset and shows that the scale of the corpus can be used to achieve SOTA representations with a simple recipe.
35
- \n\nALIGN is not open-sourced and the `kakaobrain/align-base` model used for this demo is based on the Kakao Brain implementation that follows the original paper. The model is trained on the open source [COYO](https://github.com/kakaobrain/coyo-dataset) dataset by the Kakao Brain team.
36
  To perform zero-shot image classification with ALIGN, upload an image and enter your candidate labels as free-form text separated by a comma followed by a space.</p>
37
  </div>
38
  </div>
 
29
  </div>
30
  <div class="text">
31
  <p>Gradio demo for <a href="https://huggingface.co/docs/transformers/main/en/model_doc/align">ALIGN</a>,
32
+ as introduced in <a href="https://arxiv.org/abs/2102.05918"></a><i>"Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision"</i>. ALIGN features a dual-encoder architecture with EfficientNet and BERT as its text and vision encoders, and learns to align visual and text representations with contrastive learning.
 
33
  Unlike previous work, ALIGN leverages a massive noisy dataset and shows that the scale of the corpus can be used to achieve SOTA representations with a simple recipe.
34
+ \n\nALIGN is not open-sourced and the `kakaobrain/align-base` model used for this demo is based on the Kakao Brain implementation that follows the original paper. The model is trained on the open source [COYO](https://github.com/kakaobrain/coyo-dataset) dataset by the Kakao Brain team.
35
  To perform zero-shot image classification with ALIGN, upload an image and enter your candidate labels as free-form text separated by a comma followed by a space.</p>
36
  </div>
37
  </div>