Simplify zoom workflow and prefer ImageEditor hand/pan path

app.py

@@ -52,6 +52,7 @@ model = AutoModel.from_pretrained(MODEL_NAME, _attn_implementation=_attn_impl, t
 BASE_SIZE = 1024
 IMAGE_SIZE = 768
 CROP_MODE = True
+WORKSPACE_DEFAULT_SCALE = 89
 GROUNDING_PATTERN = re.compile(r'<\|ref\|>(.*?)<\|/ref\|><\|det\|>(.*?)<\|/det\|>', re.DOTALL)
 INFER_DEBUG_FILTERS = ['PATCHES', '====', 'BASE:', 'directly resize', 'NO PATCHES', 'torch.Size', '%|']
 EQUATION_ZOOM_PROMPT = "<image>\n<|grounding|>Locate each individual equation or math line."
@@ -1352,7 +1353,7 @@ def _scale_workspace_image(img, workspace_scale):
     try:
         scale = max(60, min(220, int(workspace_scale)))
     except Exception:
-        scale = 89
+        scale = WORKSPACE_DEFAULT_SCALE
     if scale == 100:
         return img
     ratio = scale / 100.0
@@ -1361,7 +1362,7 @@ def _scale_workspace_image(img, workspace_scale):
     resample = Image.Resampling.BILINEAR if hasattr(Image, "Resampling") else Image.BILINEAR
     return img.resize((new_w, new_h), resample)
 
-def load_image_with_size(file_path, page_num=1, workspace_scale=100):
+def load_image_with_size(file_path, page_num=1, workspace_scale=WORKSPACE_DEFAULT_SCALE):
     img = load_image(file_path, page_num)
     if img is None:
         return None, None, None
@@ -1444,9 +1445,9 @@ with gr.Blocks(**blocks_kwargs) as demo:
                         editor_kwargs["eraser"] = gr.Eraser(default_size=26)
                     except TypeError:
                         editor_kwargs["eraser"] = gr.Eraser()
-                if HAS_PAINT:
+                if HAS_IMAGE_EDITOR:
                     try:
-                        region_editor = gr.Paint(
+                        region_editor = gr.ImageEditor(
                             label="Image Workspace",
                             show_label=False,
                             type="pil",
@@ -1455,20 +1456,20 @@ with gr.Blocks(**blocks_kwargs) as demo:
                         )
                     except TypeError:
                         try:
-                            region_editor = gr.Paint(
+                            region_editor = gr.ImageEditor(
                                 label="Image Workspace",
                                 show_label=False,
                                 height=640,
                                 **editor_kwargs,
                             )
                         except TypeError:
-                            region_editor = gr.Paint(
+                            region_editor = gr.ImageEditor(
                                 label="Image Workspace",
                                 show_label=False,
                                 height=640,
                             )
                 else:
-                    region_editor = gr.ImageEditor(
+                    region_editor = gr.Paint(
                         label="Image Workspace",
                         show_label=False,
                         type="pil",
@@ -1483,14 +1484,6 @@ with gr.Blocks(**blocks_kwargs) as demo:
             gr.Markdown("### OCR Workflow")
             task = gr.Dropdown(list(TASK_PROMPTS.keys()), value="📋 Markdown", label="Task")
             input_scope = gr.Radio(["Entire Page", "Selected Region"], value="Entire Page", label="Input Scope")
-            workspace_scale = gr.Slider(
-                minimum=60,
-                maximum=220,
-                step=1,
-                value=89,
-                label="Workspace Scale (%)",
-                info="Use this instead of in-canvas zoom for smoother live drawing.",
-            )
 
             selection_controls = gr.Row(visible=False)
             with selection_controls:
@@ -1567,7 +1560,7 @@ with gr.Blocks(**blocks_kwargs) as demo:
         2. Choose **Input Scope**:
            - `Entire Page` for the full page.
            - `Selected Region` for a specific area.
-        2a. For small math, increase **Workspace Scale** instead of using in-canvas zoom (better live drawing feedback).
+        2a. Workspace opens at **89% scale** by default to keep small math readable while avoiding extra zoom adjustments.
         3. For `Selected Region`, use the **Image Workspace**:
            - Recommended: freehand selection (draw/highlight target); app uses an automatic bounding box around your marks.
            - Optional rectangle selection: use the **Crop** tool.
@@ -1608,16 +1601,13 @@ with gr.Blocks(**blocks_kwargs) as demo:
     task.change(select_boxes, [task], [tabs])
     input_scope.change(toggle_scope_ui, [input_scope], [workspace_hint, selection_controls, selection_status, selected_regions_gallery])
     if HAS_REGION_WORKSPACE and region_editor is not None:
-        file_in.change(load_image_with_size, [file_in, page_selector, workspace_scale], [region_editor, workspace_base_size, workspace_base_image])
-        page_selector.change(load_image_with_size, [file_in, page_selector, workspace_scale], [region_editor, workspace_base_size, workspace_base_image])
-        workspace_scale.change(load_image_with_size, [file_in, page_selector, workspace_scale], [region_editor, workspace_base_size, workspace_base_image])
+        file_in.change(load_image_with_size, [file_in, page_selector], [region_editor, workspace_base_size, workspace_base_image])
+        page_selector.change(load_image_with_size, [file_in, page_selector], [region_editor, workspace_base_size, workspace_base_image])
         region_editor.change(sync_workspace_state, [region_editor, workspace_base_image], [workspace_base_size, workspace_base_image])
         file_in.change(_reset_selected_regions, outputs=[selected_regions_state, selected_regions_gallery, selection_status])
         page_selector.change(_reset_selected_regions, outputs=[selected_regions_state, selected_regions_gallery, selection_status])
-        workspace_scale.change(_reset_selected_regions, outputs=[selected_regions_state, selected_regions_gallery, selection_status])
         file_in.change(_reset_drawn_mask, outputs=[drawn_mask_state])
         page_selector.change(_reset_drawn_mask, outputs=[drawn_mask_state])
-        workspace_scale.change(_reset_drawn_mask, outputs=[drawn_mask_state])
 
     add_region_btn.click(
         add_selected_region,