Spaces:
Sleeping
Sleeping
Update app.py
Browse filesImproved demo desc.
app.py
CHANGED
@@ -388,10 +388,9 @@ with gr.Blocks(theme=gr.themes.Soft(), css=custom_css, title="PEEB") as demo:
|
|
388 |
|
389 |
with gr.Column():
|
390 |
title_text = gr.Markdown("# A classifier with Part-based Explainable and Editable Bottleneck (PEEB) | Demo")
|
|
|
391 |
gr.Markdown(
|
392 |
"""
|
393 |
-
Here is a PEEB classifier pre-trained on Bird-11K and finetuned on CUB-200 (see our [NAACL 2024 paper](https://arxiv.org/abs/2403.05297) and [code](https://github.com/anguyen8/peeb/tree/inspect_ddp)).\n The demo shows how one runs PEEB on an existing image and edit descriptors to modify the classifier (without any re-training).
|
394 |
-
\n
|
395 |
Steps:
|
396 |
1. **Select an image**. Then, PEEB will show its grounded explanations and the top-1 predicted label with associated softmax confidence score.
|
397 |
2. **Hover mouse over text descriptors** to see the corresponding region used to match to each text descriptor.
|
|
|
388 |
|
389 |
with gr.Column():
|
390 |
title_text = gr.Markdown("# A classifier with Part-based Explainable and Editable Bottleneck (PEEB) | Demo")
|
391 |
+
gr.Markdown("Here is a PEEB classifier pre-trained on Bird-11K and finetuned on CUB-200 (see our [NAACL 2024 paper](https://arxiv.org/abs/2403.05297) and [code](https://github.com/anguyen8/peeb/tree/inspect_ddp)).\n The demo shows how one runs PEEB on an existing image and edit descriptors to modify the classifier (without any re-training).")
|
392 |
gr.Markdown(
|
393 |
"""
|
|
|
|
|
394 |
Steps:
|
395 |
1. **Select an image**. Then, PEEB will show its grounded explanations and the top-1 predicted label with associated softmax confidence score.
|
396 |
2. **Hover mouse over text descriptors** to see the corresponding region used to match to each text descriptor.
|