Merge pull request #7 from yvrjsharma/add-gradio-demo

added gradio demo and instructions to run it in the readme

README.md

@@ -91,6 +91,20 @@ model, processor = load_model(), load_processor()
 predictions = batch_inference([image], model, processor)
 ```
 
+### Gradio Demo
+
+Install Gradio
+
+```python
+pip install gradio
+```
+
+Run the app gradio_demo.py
+
+```python
+python gradio_demo.py
+```
+
 ## Text recognition
 
 Coming soon.

gradio_demo.py

@@ -0,0 +1,57 @@
+import gradio as gr
+from PIL import Image, ImageDraw
+from surya.detection import batch_inference
+from surya.model.segformer import load_model, load_processor
+
+model, processor = load_model(), load_processor()
+
+def surya(img):
+
+    # surya predictions is a list of dicts for the given image
+    predictions = batch_inference([img], model, processor)
+    bboxes = predictions[0]['bboxes']
+    vertical_lines = predictions[0]['vertical_lines']
+    horizontal_lines = predictions[0]['horizontal_lines']
+
+    # Initialize the drawing context with the image as background
+    draw = ImageDraw.Draw(img)
+
+    # OCR predictions (replace the sample data with your actual OCR output)
+    predictions = {
+        'bboxes': bboxes, # bounding boxes data here
+        'vertical_lines':  vertical_lines, # vertical lines data here
+        'horizontal_lines':  horizontal_lines # your horizontal lines data here
+    }
+
+    # Draw bounding boxes
+    for bbox in predictions['bboxes']:
+        draw.rectangle(bbox, outline='red', width=2)
+
+    # Draw vertical lines
+    for vline in predictions['vertical_lines']:
+        x1, y1, x2, y2 = vline['bbox']
+        draw.line((x1, y1, x2, y2), fill='blue', width=2)
+
+    # Draw horizontal lines
+    for hline in predictions['horizontal_lines']:
+        x1, y1, x2, y2 = hline['bbox']
+        draw.line((x1, y1, x2, y2), fill='green', width=2)
+
+    # return the final image 
+    return img 
+
+# Blocks API
+with gr.Blocks() as demo:
+  # title for the app
+  gr.HTML("<h1><center> SURYA Demo </h1></center>")
+  # input image component
+  input_image = gr.Image(label="Input Image", type='pil')
+  # run inference on the input image
+  btn = gr.Button("Run Surya")
+  # output image component
+  output_image = gr.Image(label="Surya Output")
+  btn.click(fn=surya, inputs=input_image, outputs=output_image, api_name="surya")
+
+
+if __name__ == "__main__":
+    demo.launch()