nielsr HF Staff Claude commited on
Commit
680f872
·
1 Parent(s): c27b8fb

Add example images and questions for better UX

Browse files

- Add receipt example image to all three tabs for easy testing
- Include sample questions for the Chat tab to guide users
- Users can now click examples to quickly try different features

Examples:
- Image: Official KOSMOS-2.5 receipt sample
- Questions: Total amount, items purchased, receipt date, subtotal

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <[email protected]>

Files changed (1) hide show
  1. app.py +16 -0
app.py CHANGED
@@ -201,6 +201,10 @@ with gr.Blocks(title="KOSMOS-2.5 Document AI Demo", theme=gr.themes.Soft()) as d
201
  with gr.Row():
202
  with gr.Column():
203
  md_image = gr.Image(type="pil", label="Upload Document Image")
 
 
 
 
204
  md_button = gr.Button("Generate Markdown", variant="primary")
205
  with gr.Column():
206
  md_output = gr.Textbox(
@@ -215,6 +219,10 @@ with gr.Blocks(title="KOSMOS-2.5 Document AI Demo", theme=gr.themes.Soft()) as d
215
  with gr.Row():
216
  with gr.Column():
217
  ocr_image = gr.Image(type="pil", label="Upload Document Image")
 
 
 
 
218
  ocr_button = gr.Button("Extract Text with Coordinates", variant="primary")
219
  with gr.Column():
220
  with gr.Row():
@@ -230,11 +238,19 @@ with gr.Blocks(title="KOSMOS-2.5 Document AI Demo", theme=gr.themes.Soft()) as d
230
  with gr.Row():
231
  with gr.Column():
232
  chat_image = gr.Image(type="pil", label="Upload Document Image")
 
 
 
 
233
  chat_question = gr.Textbox(
234
  label="Ask a question about the document",
235
  placeholder="e.g., What is the total amount on this receipt?",
236
  lines=2
237
  )
 
 
 
 
238
  chat_button = gr.Button("Get Answer", variant="primary")
239
  with gr.Column():
240
  chat_output = gr.Textbox(
 
201
  with gr.Row():
202
  with gr.Column():
203
  md_image = gr.Image(type="pil", label="Upload Document Image")
204
+ gr.Examples(
205
+ examples=["https://huggingface.co/microsoft/kosmos-2.5/resolve/main/receipt_00008.png"],
206
+ inputs=md_image
207
+ )
208
  md_button = gr.Button("Generate Markdown", variant="primary")
209
  with gr.Column():
210
  md_output = gr.Textbox(
 
219
  with gr.Row():
220
  with gr.Column():
221
  ocr_image = gr.Image(type="pil", label="Upload Document Image")
222
+ gr.Examples(
223
+ examples=["https://huggingface.co/microsoft/kosmos-2.5/resolve/main/receipt_00008.png"],
224
+ inputs=ocr_image
225
+ )
226
  ocr_button = gr.Button("Extract Text with Coordinates", variant="primary")
227
  with gr.Column():
228
  with gr.Row():
 
238
  with gr.Row():
239
  with gr.Column():
240
  chat_image = gr.Image(type="pil", label="Upload Document Image")
241
+ gr.Examples(
242
+ examples=["https://huggingface.co/microsoft/kosmos-2.5/resolve/main/receipt_00008.png"],
243
+ inputs=chat_image
244
+ )
245
  chat_question = gr.Textbox(
246
  label="Ask a question about the document",
247
  placeholder="e.g., What is the total amount on this receipt?",
248
  lines=2
249
  )
250
+ gr.Examples(
251
+ examples=["What is the total amount on this receipt?", "What items were purchased?", "When was this receipt issued?", "What is the subtotal?"],
252
+ inputs=chat_question
253
+ )
254
  chat_button = gr.Button("Get Answer", variant="primary")
255
  with gr.Column():
256
  chat_output = gr.Textbox(