First commit

.gitignore

@@ -0,0 +1 @@
+__pycache__

__init__.py

@@ -0,0 +1,6 @@
+from .depth_background_blur import DepthBackgroundBlur
+
+NODE_CLASS_MAPPINGS = {"DepthBackgroundBlur": DepthBackgroundBlur}
+NODE_DISPLAY_NAME_MAPPINGS = {"DepthBackgroundBlur": "DepthBackgroundBlur"}
+
+__all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS"]

depth_background_blur.py

@@ -0,0 +1,56 @@
+import torch
+import cv2
+import numpy as np
+
+
+class DepthBackgroundBlur:
+    CATEGORY = "background-edit"
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "execute"
+
+    @classmethod
+    def INPUT_TYPES(s):
+        return {
+            "required": {
+                "images": ("IMAGE",),
+                "depth_maps": ("IMAGE",),
+                "blur_strength": ("INT", {"default": 51}),
+                "threshold": ("INT", {"default": 125}),
+            }
+        }
+
+    def execute(
+        self,
+        images: torch.Tensor,
+        depth_maps: torch.Tensor,
+        blur_strength: int,
+        threshold: int,
+    ):
+        if images.shape[0] != depth_maps.shape[0]:
+            raise Exception("mismatch number of images and depth maps")
+
+        images_np = (images * 255.0).clamp(0, 255).to(dtype=torch.uint8).cpu().numpy()
+        depth_maps_np = (
+            (depth_maps * 255.0).clamp(0, 255).to(dtype=torch.uint8).cpu().numpy()
+        )
+
+        results = []
+        for image, depth_map in zip(images_np, depth_maps_np):
+            # Convert to BGR for OpenCV
+            image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
+            binary_mask = (depth_map > threshold).astype(np.uint8) * 255
+            blurred = cv2.GaussianBlur(image, (blur_strength, blur_strength), 0)
+
+            binary_mask_3ch = binary_mask / 255.0
+            inverse_mask_3ch = 1.0 - binary_mask_3ch
+
+            foreground = image * binary_mask_3ch
+            background = blurred * inverse_mask_3ch
+            result = cv2.add(foreground, background)
+
+            # Convert to RGB for torch
+            result = cv2.cvtColor(result.astype(np.uint8), cv2.COLOR_BGR2RGB)
+
+            results.append(result)
+
+        return (torch.from_numpy(np.stack(results, axis=0).astype(np.float32) / 255.0),)

requirements.txt

@@ -0,0 +1,3 @@
+opencv-python
+numpy
+torch