ComfyUI_depthMapOperation

ComfyUI_depthMapOperation

A simple set of nodes to generate a point cloud from an image and its depth map, perform transformations and some basic operations.

Here an example of what can be done (not the gif, but the various rotations)

<div class="grid" align="center" markdown> <img src=".//assets/start.jpg" width=40%> <img src=".//assets/gif.gif" width=40%> </div>

---

Installation

On the console in the custom_nodes filder execute:

git clone https://github.com/chri002/ComfyUI_depthMapOperation

Requirements

these nodes require in order to function:

• torch
• numpy
• opencv-python
• scipy
• pandas

Workflow

---

Nodes

Image To Points (Torch)

Description

GPU-accelerated version using PyTorch tensors. Maintains gradient flow and supports automatic device placement.

Input Parameters:

• image: Input RGB/RGBA image
• depth_image: Depth map image
• depth: Z-axis scaling factor (1-1024)
• quality: Downsampling quality (1=1 point:1 pixel , 16=16 interpolate points every 2 pixel)

Output:

• Points3D: XYZ coordinates + RGB colors

---

Transform Points

Description

Applies 3D transformations to point clouds (rotation, translation, scaling).

Input Parameters:

• points: Input point cloud (Points3D)
• rot_x/y/z: Euler angles in degrees
• trl_x/y/z: Translation offsets
• scale_x/y/z: Axis-specific scaling factors

Output:

• Points3D: Transformed point cloud (XYZ coordinates + RGB colors)

---

Points To Image (Orthographic)

Description

Renders 3D points to 2D image using orthographic projection.

Input Parameters:

• images: Template for output dimensions
• points: Point cloud to render (Points3D)
• color: Enable RGB coloring

Output:

• IMAGE: Rendered grayscale/RGB image

---

Points To Image (Projection)

Description

Perspective projection renderer with customizable FOV.

Input Parameters:

• images: Template for output dimensions
• points: Point cloud to render (Points3D)
• color: Enable RGB coloring
• fov: Field of View in degrees (1-2000)

Output:

• IMAGE: Rendered grayscale/RGB image

---

Cube Limit

Description

Filters points within relative cube dimensions (0-100% of original bounds).

Input Parameters:

• points: Point cloud to render (Points3D)
• 6 axis range parameters (x_min-x_max, etc.)

Output:

• Points3D: Subset of points within cube (XYZ coordinates + RGB colors)

---

Clean Points (KDTree)

Description

Removes outliers using KDTree neighborhood analysis.

Parameters:

• points: Point cloud to render (Points3D)
• k: Minimum neighbors required
• m: Max neighbor distance threshold

Output:

• Points3D: Cleaned point cloud (XYZ coordinates + RGB colors)

---

Interpolate Points (KDTree)

Description

Generates new points through neighborhood-based interpolation using KDTree. Enhances point cloud density in sparse regions by creating intermediate points between existing neighbors.

Input Parameters:

• points: Input 3D point cloud
• value: (0-1) Blend ratio for new points (0=keep original, 1=full interpolation)
• n: Number of nearest neighbors to consider (0-32)

Output:

• Points3D: Point cloud with added interpolated points (XYZ coordinates + RGB colors)

---

Export To PLY

Description

Exports point cloud to PLY format (ASCII/binary).

Input Parameters:

• points: Point cloud to render (Points3D)
• multiple_files: Split XYZ/RGB data
• format_out: File encoding format

---

Import PLY

Description

Import PLY point cloud files into compatible Point3D format. (Sperimental)

Input Parameter:

• .ply file selection

Output:

• Points3D: Loaded point cloud data (XYZ coordinates + RGB colors)

---

Cloud Points Info

Description

Displays point cloud statistics and coordinate ranges.

Output:

• STRING: Formatted summary text

---

DEMO FUNCTIONS

<a target="_blank" href="https://colab.research.google.com/github/chri002/ComfyUI_depthMapOperation/blob/main/test.ipynb"> <img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Test In Colab" height="30px"/> </a>

TODO

• Fix artifacts with some extreme values