ComfyUI Node
MotionMaskFromDepth
Detects dynamic pixels from a depth+pose sequence (1.0 = moving).
MotionMaskFromDepth
- depth_seq
- trajectory
- motion_mask
◄input_projection▾►
◄input_horizontal_fov90.00►
◄threshold0.10►
◄frame_gap4►
◄dilate2►
◄deviceauto►
CategoryCamera/GSplat4D
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| depth_seq | TENSOR | Depth sequence [T,H,W] (radial distance). | |
| trajectory | TENSOR | World-to-camera poses [T,4,4] (or [K,4,4]; interpolated to T). | |
| input_projection | COMBO | 3 options: PINHOLE, FISHEYE, EQUIRECTANGULAR | |
| input_horizontal_fov | FLOAT | 90.001–360 | — |
| threshold | FLOAT | 0.100–10 | Relative depth residual |d_proj - d_sampled|/d_sampled above which a pixel is dynamic. |
| frame_gap | INT | 41–256 | Temporal gap for forward/backward reprojection checks. |
| dilate | INT | 20–64 | Dilation radius (pixels) of the dynamic mask. |
| deviceopt | COMBO | auto | 3 options: auto, cpu, cuda |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| motion_mask | MASK | — |