Update app.py

app.py

@@ -59,24 +59,38 @@ def run_ocrd_pipeline(img_path, status=gr.Progress(), binarize_mode='detailed',
 
 
 demo_data = [
-    './demo_data/act_image.jpg',
-    './demo_data/newjersey2_image.jpg',
-    './demo_data/washington_image.jpg'
+    ['../demo_data/act_image.jpg', None],
+    ['../demo_data/newjersey2_image.jpg', None],
+    ['../demo_data/washington_image.jpg', None]
 ]
 
 
-iface = gr.Interface(run_ocrd_pipeline,
+description = """<ul>
+                    <li>This interactive demo showcases an 'Optical Character Recognition Digitization' pipeline that processes images to recognize text.</li>
+                    <li>Steps include:
+                        <ol>
+                            <li>Image binarization</li>
+                            <li>Text line segmentation</li>
+                            <li>Text line extraction, filtering and deskewing</li>
+                            <li>OCR on textlines</li>
+                            <li>Printing recognized text on generated image for visualization</li>
+                        </ol>
+                    </li>
+                    <li>Optimized for <b>English</b>; other languages (e.g., German) may require OCR model fine-tuning.</li>
+                    <li>Uses free CPU-based compute, which is rather <b>slow</b>. A pipeline run will take up to 10 minutes.</li> 
+                    <li>For lengthy waits, click on the pre-computed examples below or look at example results at: <a href='https://github.com/pluniak/ocrd/tree/main/src/demo_data'>https://github.com/pluniak/ocrd/tree/main/src/demo_data</a></li>
+                    <li>The demo is based on code from my GitHub repository: <a href='https://github.com/pluniak/ocrd'>https://github.com/pluniak/ocrd</a></li>
+                    <li>Note: The demo is just a <b>first prototype</b>! OCR performance and computation speed should be optimized.</li>
+                </ul>"""
+
+iface = gr.Interface(fn = run_ocrd_pipeline,
                      title="OCRD Pipeline",
-                     description="<ul><li>This interactive demo showcases an 'Optical Character Recognition Digitization' pipeline that processes \
-                                  images to recognize text.</li> \
-                                  <li>Steps include binarization, text line segmentation, extraction, filtering and deskewing as well as OCR. \
-                                  Results are displayed as a generated overlay image.</li> \
-                                  <li>Optimized for English; other languages (e.g. German) may require OCR model fine-tuning.</li> \
-                                  <li>Uses free CPU-based compute, which is rather slow. A pipeline run will take up to 10 minutes. \
-                                  For lengthy waits, pre-computed demo results are available for download: https://github.com/pluniak/ocrd/tree/main/src/demo_data.</li> \
-                                  <li>Note: The demo is just a first version! OCR performance and computation speed can be optimized.</li> \
-                                  <li>The demo is based on code from my GitHub repository: https://github.com/pluniak/ocrd/tree/main</li></ul>",
-                     inputs=[gr.Image(type='filepath', label='Input image')],
+                     description=description,
+                     inputs=[
+                         gr.Image(type='filepath', label='Input image'), 
+                         gr.Dropdown(choices=['small', 'medium', 'large', 'adjusted'], label='Output image font size', value='small', 
+                                     info='"adjusted" will try to mimic font sizes from the input image')
+                     ],
                      outputs=gr.Image(label='Output image: overlay with recognized text', type='pil', format='jpeg'),
                      examples=demo_data)
 iface.launch()