Enlarge the images to click on

click3r/demo/app.py

@@ -64,6 +64,7 @@ from click3r.demo.views import prepare_model_view
 from click3r.demo.results import ReconstructionResult
 
 DEBUG = False
+DISPLAY_LONG_SIDE = 640  # Pixels; small inputs would otherwise be hard to click.
 # Optional comparison models are imported lazily by their load functions.
 VGGT_MODEL_NAME = "facebook/VGGT-1B"
 PI3_MODEL_NAME = "yyfz233/Pi3X"
@@ -1260,6 +1261,12 @@ def get_pair_color_rgba(pair_index):
     return [int(c * 255) for c in rgba]
 
 
+def display_size(image):
+    """Size an image is shown at; clicks are stored as normalized coordinates."""
+    scale = DISPLAY_LONG_SIDE / max(image.size)
+    return round(image.width * scale), round(image.height * scale)
+
+
 def add_crop_region_overlay(image, target_size=224):
     """
     Add semi-transparent overlay to show regions that will be cropped out.
@@ -1275,7 +1282,7 @@ def add_crop_region_overlay(image, target_size=224):
     if image is None:
         return None
 
-    img_copy = image.copy()
+    img_copy = image.resize(display_size(image), Image.LANCZOS)
     W, H = img_copy.size
 
     # Calculate resize and crop according to DUSt3R's logic
@@ -1657,7 +1664,7 @@ def handle_img1_click(evt: gr.SelectData):
 
         # Convert to normalized coordinates
         if _session()["current_image_pair"][0]:
-            width, height = _session()["current_image_pair"][0].size
+            width, height = display_size(_session()["current_image_pair"][0])
             x_norm = x_pixel / width
             y_norm = y_pixel / height
             debug_log(
@@ -1712,7 +1719,7 @@ def handle_img2_click(evt: gr.SelectData):
 
         # Convert to normalized coordinates
         if _session()["current_image_pair"][1]:
-            width, height = _session()["current_image_pair"][1].size
+            width, height = display_size(_session()["current_image_pair"][1])
             x_norm = x_pixel / width
             y_norm = y_pixel / height
             debug_log(