ComfyUI nodes to use Depth Anything V3 - the latest depth estimation model from ByteDance. For now monocular depth, camera pose estimation and 3D point clouds/gaussians are supported. Models autodownload from HuggingFace (depth-anything org). This is a first draft, let me know if you have any feedback! :)

ComfyUI-DepthAnythingV3


Fuse multi-view depth maps into a single world-space point cloud.

Uses predicted camera poses (extrinsics) to transform each view's depth
into a common world coordinate system, then combines all points.

Inputs:
- depths: Batch of depth maps [N, H, W, 3] from Multi-View 3D node
- images: Original images [N, H, W, 3] for RGB colors
- extrinsics: Camera poses JSON from Multi-View node
- intrinsics: Camera intrinsics JSON from Multi-View node
- confidence: Optional confidence maps to filter low-confidence points
- sky_mask: Optional sky segmentation to exclude sky pixels from point cloud
- use_icp: Refine alignment with ICP (slower but potentially more accurate)

Output: Single combined POINTCLOUD in world space.

Note: Requires Main series or Nested model (with camera pose prediction).
Mono/Metric models don't predict camera poses.


ComfyUI Node: DA3 Multi-View Point Cloud

Category

Inputs

Outputs

Extension: ComfyUI-DepthAnythingV3

Run ComfyUI workflows in the Cloud!