ComfyUI Node
BD Face Landmarks
MediaPipe FaceLandmarker (Tasks API) on a batch of head photos. Outputs: - LANDMARKS_BATCH consumed by BD_FlameFit. Each view carries 478 2D+3D landmarks, a 4x4 head-pose transform, and a coarse view-hint (front / left / right / rear). - Debug overlay image with landmarks + view label per view. Failed detections (rear / extreme profile) stay in the batch with detected=False so view ordering is preserved. Model bundle (~3.6MB) is searched in ComfyUI model directories under mediapipe/face_landmarker.task; auto-downloaded on first use.
BD Face Landmarks
- images
- landmarks_batch
- overlay
- status
◄min_detection_confidence0.40►
◄min_presence_confidence0.40►
◄model_path►
Category🧠BrainDead/FaceWrap
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | Batch of head photos. For the 4-view face-wrap stack front/left/right/rear in that order. | |
| min_detection_confidence | FLOAT | 0.400–1 | Lower = more profile views accepted. 0.4 is a decent profile floor. |
| min_presence_confidence | FLOAT | 0.400–1 | MediaPipe landmark-presence threshold. |
| model_pathopt | STRING | Override path to face_landmarker.task. Empty = search ComfyUI model dirs for mediapipe/face_landmarker.task (auto-download if missing). |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| landmarks_batch | BD_LANDMARKS_BATCH | — |
| overlay | IMAGE | — |
| status | STRING | — |