loading images

app.py

@@ -0,0 +1,91 @@
+import traceback
+import gradio as gr
+from utils.get_RGB_image import get_RGB_image, is_online_file, steam_online_file
+from pdf2image import convert_from_path, convert_from_bytes
+
+def similarity_fn(document_image_1, document_image_2):
+    xxx = 'BOTH' if document_image_1 and document_image_2 else 'ONE' if document_image_1 or document_image_2 else 'NONE'
+    return xxx
+
+def load_image(filename, page = 0):
+    try:
+        image = None
+        try:
+            if (is_online_file(filename)):
+                image = get_RGB_image(convert_from_bytes(steam_online_file(filename))[page])
+            else:
+                image = get_RGB_image(convert_from_path(filename)[page])
+        except:
+            image = get_RGB_image(filename)
+        return [
+            gr.Image(value=image, visible=True), 
+            None
+        ]
+    except:
+        error = traceback.format_exc()
+        return [None, gr.HTML(value=error, visible=True)]
+    
+def preview_url(url, page = 0):
+    [image, error] = load_image(url, page = page)
+    if image:
+        return [gr.Tabs(selected=0), image, error]
+    else:
+        return [gr.Tabs(selected=1), image, error]
+    
+def document_view(document_number: int):
+    gr.HTML(value=f'<h4>Load the {"first" if document_number == 1 else "second"} PDF or Document Image<h4>', elem_classes=['center'])
+    with gr.Tabs() as document_tabs:
+        with gr.Tab("From Image", id=0):
+            document = gr.Image(type="pil", label=f"Document {document_number}", visible=False)
+            document_error_message = gr.HTML(label="Error Message", visible=False)
+            document_preview = gr.UploadButton(
+                "Click to PDF or Document Image", 
+                file_types=["image", ".pdf"], 
+                file_count="single")
+        with gr.Tab("From URL", id=1):
+            document_url = gr.Textbox(
+                label=f"Document {document_number} URL",
+                info="Paste a Link/URL to PDF or Document Image",
+                placeholder="https://datasets-server.huggingface.co/.../image.jpg")
+            document_url_error_message = gr.HTML(label="Error Message", visible=False)
+            document_url_preview = gr.Button(value="Preview", variant="primary")
+    document_preview.upload(
+         fn = lambda file: load_image(file.name), 
+         inputs = [document_preview], 
+         outputs = [document, document_error_message])
+    document_url_preview.click(
+        fn = preview_url, 
+        inputs = [document_url], 
+        outputs = [document_tabs, document, document_url_error_message])
+    return document
+
+def app():
+    title = 'Document Similarity Search Using Visual Layout Features'
+    description = "<h2>Document Similarity Search using Detectron2<h2>"
+    article = "<h4>More details, Links about this! - Document Similarity Search using Detectron2<h4>"
+    css = '''
+    image { max-height="86vh" !important; }
+    .center { display: flex; flex: 1 1 auto; align-items: center; align-content: center; justify-content: center; justify-items: center; }
+  '''
+    with gr.Blocks(title=title, css=css) as app:
+        with gr.Row():
+            gr.HTML(value=description, elem_classes=['center'])
+        with gr.Row(equal_height = False):
+            with gr.Column():
+                document_1_image = document_view(1)
+            with gr.Column():
+                document_2_image = document_view(2)
+        with gr.Row():
+            # gr.HTML(value=article, elem_classes=['center'])
+            submit = gr.Button(value="Preview", variant="primary")
+            similarity_output = gr.HTML(value=article, elem_classes=['center'], visible=False)
+            submit.click(
+                fn=lambda image: similarity_fn(image, just_image=False),
+                inputs=[document_1_image, document_2_image],
+                outputs=[similarity_output])
+        
+        # threshold.change(
+        #     fn=lambda image, threshold: similarity_fn(image, threshold, just_image=False),
+        #     inputs=[document_1, threshold],
+        #     outputs=[annotated_document_image, message])
+    return app.launch(debug=True)

utils/get_RGB_image.py

@@ -0,0 +1,18 @@
+from PIL import Image
+from urllib.parse import urlparse
+import requests
+
+def is_online_file(url: str) -> bool:
+    return urlparse(url).scheme in ["http", "https"]
+
+def steam_online_file(url: str) -> bytes:
+    return requests.get(url, stream=True).raw
+
+def get_RGB_image(image_or_path: str | Image.Image) -> bytes:
+    if isinstance(image_or_path, str):
+        if is_online_file(image_or_path):  # Online
+            content = steam_online_file(image_or_path)
+            image_or_path = Image.open(content)
+        else:  # Local
+            image_or_path = Image.open(image_or_path)
+    return image_or_path.convert("RGB")