output_document_image_1_hash

analysis.ipynb

@@ -301,14 +301,14 @@
     },
     {
       "cell_type": "code",
-      "execution_count": 33,
+      "execution_count": 21,
       "metadata": {},
       "outputs": [
         {
           "name": "stdout",
           "output_type": "stream",
           "text": [
-            "Running on local URL:  http://127.0.0.1:7860\n",
+            "Running on local URL:  http://127.0.0.1:7861\n",
             "\n",
             "To create a public link, set `share=True` in `launch()`.\n"
           ]
@@ -316,7 +316,7 @@
         {
           "data": {
             "text/html": [
-              "<div><iframe src=\"http://127.0.0.1:7860/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
+              "<div><iframe src=\"http://127.0.0.1:7861/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
             ],
             "text/plain": [
               "<IPython.core.display.HTML object>"
@@ -329,14 +329,6 @@
           "name": "stdout",
           "output_type": "stream",
           "text": [
-            "{'predicted_bboxes': 'predicted_bboxes', 'predicted_scores': 'predicted_scores', 'predicted_labels': 'predicted_labels'}\n",
-            "document_image_1.info.get(annotation_key) == True, start: False\n",
-            "document_image_1.info.get(annotation_key) == True, middle: False\n",
-            "document_image_1.info.get(annotation_key) == True, end: True\n",
-            "{'predicted_bboxes': 'reduced_predicted_bboxes', 'predicted_scores': 'reduced_predicted_scores', 'predicted_labels': 'reduced_predicted_labels'}\n",
-            "document_image_1.info.get(annotation_key) == True, start: False\n",
-            "document_image_1.info.get(annotation_key) == True, middle: False\n",
-            "document_image_1.info.get(annotation_key) == True, end: True\n",
             "Keyboard interruption in main thread... closing server.\n"
           ]
         },
@@ -344,7 +336,7 @@
           "data": {
             "text/plain": []
           },
-          "execution_count": 33,
+          "execution_count": 21,
           "metadata": {},
           "output_type": "execute_result"
         }
@@ -356,6 +348,26 @@
         "config_path = '../detectron2-layout-parser/config.yaml'\n",
         "app(model_path=model_path, config_path=config_path, debug=True)"
       ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": 15,
+      "metadata": {},
+      "outputs": [
+        {
+          "data": {
+            "text/plain": [
+              "False"
+            ]
+          },
+          "execution_count": 15,
+          "metadata": {},
+          "output_type": "execute_result"
+        }
+      ],
+      "source": [
+        "'3d29390000ffff3b' == '3d29090100ffff39'"
+      ]
     }
   ],
   "metadata": {

main.py

@@ -13,10 +13,12 @@ label_map = {0: 'Caption', 1: 'Footnote', 2: 'Formula', 3: 'List-item', 4: 'Page
 label_names = list(label_map.values())
 color_map = {'Caption': '#acc2d9', 'Footnote': '#56ae57', 'Formula': '#b2996e', 'List-item': '#a8ff04', 'Page-footer': '#69d84f', 'Page-header': '#894585', 'Picture': '#70b23f', 'Section-header': '#d4ffff', 'Table': '#65ab7c', 'Text': '#952e8f', 'Title': '#fcfc81'}
 cache = {
-    'document_image_1_hash': None,
-    'document_image_2_hash': None,
+    'output_document_image_1_hash': None,
+    'output_document_image_2_hash': None,
     'document_image_1_features': None,
     'document_image_2_features': None,
+    'original_document_image_1': None,
+    'original_document_image_2': None
 }
 pre_message_style = 'overflow:auto;border:2px solid pink;padding:4px;border-radius:4px;'
 visualize_bboxes_on_image_kwargs = {
@@ -27,7 +29,7 @@ visualize_bboxes_on_image_kwargs = {
     'label_rectangle_left_margin': 0,
     'label_rectangle_top_margin': 0
 }
-vectors_types = ['vectors', 'weighted_vectors', 'reduced_vectors', 'reduced_weighted_vectors']
+vectors_types = ['vectors', 'weighted_vectors', 'reduced_vectors', 'weighted_reduced_vectors']
 
 annotation_key = 'is_annotated_document_image'
 annotation_original_image_key = 'original_image'
@@ -55,25 +57,24 @@ def similarity_fn(model: lp.Detectron2LayoutModel, document_image_1: Image.Image
         if document_image_1 is None or document_image_2 is None:
             message = f'<pre style="{pre_message_style}">Please load both the documents to compare.<pre>'
         else:
-            document_image_1 = get_original_document_image(document_image_1)
-            document_image_2 = get_original_document_image(document_image_2)
+            input_document_image_1_hash = str(average_hash(document_image_1))
+            input_document_image_2_hash = str(average_hash(document_image_2))
 
-            document_image_1_hash = str(average_hash(document_image_1))
-            document_image_2_hash = str(average_hash(document_image_2))
-
-            if document_image_1_hash == cache['document_image_1_hash']:
+            if input_document_image_1_hash == cache['output_document_image_1_hash']:
                 document_image_1_features = cache['document_image_1_features']
+                document_image_1 = cache['original_document_image_1']
             else:
                 document_image_1_features = get_features(document_image_1, model, label_names)
-                cache['document_image_1_hash'] = document_image_1_hash
                 cache['document_image_1_features'] = document_image_1_features
+                cache['original_document_image_1'] = document_image_1
 
-            if document_image_2_hash == cache['document_image_2_hash']:
+            if input_document_image_2_hash == cache['output_document_image_2_hash']:
                 document_image_2_features = cache['document_image_2_features']
+                document_image_2 = cache['original_document_image_2']
             else:
                 document_image_2_features = get_features(document_image_2, model, label_names)
-                cache['document_image_2_hash'] = document_image_2_hash
                 cache['document_image_2_features'] = document_image_2_features
+                cache['original_document_image_2'] = document_image_2
 
             [[similarity]] = cosine_similarity(
                 [
@@ -83,23 +84,25 @@ def similarity_fn(model: lp.Detectron2LayoutModel, document_image_1: Image.Image
                     cache['document_image_2_features'][vectors_type]
                 ])
             message = f'<pre style="{pre_message_style}">Similarity between the two documents is: {round(similarity, 4)}<pre>'
-            annotated_document_image_1 = visualize_bboxes_on_image(
+            document_image_1 = visualize_bboxes_on_image(
                 image = document_image_1,
                 bboxes = cache['document_image_1_features'][annotations['predicted_bboxes']],
                 titles = [f'{label}, score:{round(score, 2)}' for label, score in zip(
                     cache['document_image_1_features'][annotations['predicted_labels']], 
                     cache['document_image_1_features'][annotations['predicted_scores']])],
                 **visualize_bboxes_on_image_kwargs)
-            annotated_document_image_2 = visualize_bboxes_on_image(
+            document_image_2 = visualize_bboxes_on_image(
                 image = document_image_2,
                 bboxes = cache['document_image_2_features'][annotations['predicted_bboxes']],
                 titles = [f'{label}, score:{score}' for label, score in zip(
                     cache['document_image_2_features'][annotations['predicted_labels']], 
                     cache['document_image_2_features'][annotations['predicted_scores']])],
                 **visualize_bboxes_on_image_kwargs)
+            
+            cache['output_document_image_1_hash'] = str(average_hash(document_image_1))
+            cache['output_document_image_2_hash'] = str(average_hash(document_image_2))
+
             show_vectors_type = True
-            document_image_1 = annotate_document_image(annotated_document_image_1, document_image_1)
-            document_image_2 = annotate_document_image(annotated_document_image_2, document_image_2)
     except Exception as e:
         message = f'<pre style="{pre_message_style}">{traceback.format_exc()}<pre>'
     return [
@@ -138,7 +141,7 @@ 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 = gr.Image(type="pil", label=f"Document {document_number}", visible=False, interactive=False, show_download_button=False)
             document_error_message = gr.HTML(label="Error Message", visible=False)
             document_preview = gr.UploadButton(
                 "Upload PDF or Document Image", 

utils/get_features.py

@@ -102,7 +102,7 @@ def get_features(image: Image.Image, model: lp.Detectron2LayoutModel, label_name
     weighted_jaccard_index = False,
     **reduced_predictions)
   
-  reduced_weighted_vectors = get_vectors(
+  weighted_reduced_vectors = get_vectors(
     sub_images_bboxes = sub_images_bboxes,
     label_names = label_names,
     weighted_jaccard_index = True,
@@ -119,5 +119,5 @@ def get_features(image: Image.Image, model: lp.Detectron2LayoutModel, label_name
       'reduced_predicted_scores': reduced_predictions['predicted_scores'],
       'reduced_predicted_labels': reduced_predictions['predicted_labels'],
       'reduced_vectors': list(reduced_vectors),
-      'weighted_reduced_vectors': list(reduced_weighted_vectors),
+      'weighted_reduced_vectors': list(weighted_reduced_vectors),
   }