Fix overlap detection with base_mask: include base_mask in initial canvas for proper overlap checking

nodes.py

@@ -367,15 +367,15 @@ class AdvancedShapePlacementNode(comfy_io.ComfyNode):
             max_consecutive_fails=max_consecutive_fails,
             progress_callback=progress_callback,
             debug_mode=bool(debug_mode),
+            base_canvas=base_np,
         )
         
-        # Store the added mask (before combining with base)
-        added_mask = composite_mask.copy()
-        
-        # Combine with base mask if provided
+        # Calculate added mask (composite_mask includes base, so subtract it)
         if base_np is not None:
-            # Use maximum to combine masks (union operation)
-            composite_mask = np.maximum(composite_mask, base_np)
+            # added_mask = newly placed shapes only (composite - base)
+            added_mask = np.maximum(composite_mask - base_np, 0)
+        else:
+            added_mask = composite_mask.copy()
         
         # Convert results back to torch
         result_composite = numpy_to_torch_mask(composite_mask)

utils/placement.py

@@ -357,6 +357,7 @@ def place_shapes_in_region(
     max_consecutive_fails: int = 100,
     progress_callback = None,  # Callback for ComfyUI progress updates
     debug_mode: bool = False,
+    base_canvas: Optional[np.ndarray] = None,
 ) -> Tuple[np.ndarray, List[Dict]]:
     """
     Place multiple shape masks within a region mask with overlap control.
@@ -405,7 +406,11 @@ def place_shapes_in_region(
     region_height, region_width = region_mask.shape
     
     # Initialize output canvas
-    canvas = np.zeros_like(region_mask, dtype=np.float32)
+    # If base_canvas provided, use it as starting point (includes existing shapes for overlap detection)
+    if base_canvas is not None:
+        canvas = base_canvas.copy()
+    else:
+        canvas = np.zeros_like(region_mask, dtype=np.float32)
     
     # Store placement metadata
     placements = []
@@ -554,9 +559,9 @@ def place_shapes_in_region(
         time_checks += time.time() - t0
         
         # Calculate overlap (only within canvas bounds)
-        # Optimize: skip if no existing shapes in this region
+        # Check overlap with existing shapes (including base_canvas if provided)
         t0 = time.time()
-        if placed_count > 0 and np.any(placement_region > 0.5):
+        if np.any(placement_region > 0.5):
             overlap_area = np.sum((clipped_mask > 0.5) & (placement_region > 0.5))
             overlap_ratio_actual = overlap_area / new_mask_area
             
@@ -571,8 +576,8 @@ def place_shapes_in_region(
                 consecutive_fails += 1
                 time_overlap += time.time() - t0
                 continue
-        elif debug_mode and placed_count == 0:
-            print(f"  Overlap: SKIP (first shape)")
+        elif debug_mode:
+            print(f"  Overlap: SKIP (no existing shapes in region)")
         time_overlap += time.time() - t0
         
         # Check if mostly within valid region (using clipped mask)