Image face detect

POST /api/image-face-detect

Detect faces in an image. Returns each face's bounding rectangle plus a confidence score. Multiple faces per image are supported.

Detects faces in an image and returns: - A bounding rectangle for each face. - A confidence score (0–1). - 5 facial-landmark coordinates per face (eyes / nose / mouth corners). Multiple faces per image are supported. Pair with `image-annotate-detections` to draw the boxes onto a visualisation.

Inputs

Name	Type	Default	Description
image*	file	—	Image bytes containing zero or more faces.
scoreThreshold	number (0…1)	0.6	Per-face confidence threshold (0..1). Lower = more recall, more false positives.
nmsThreshold	number (0…1)	0.3	Non-max-suppression IoU threshold (0..1).
topK	number (1…20000)	5000	Max candidate boxes considered before NMS.
multiple	boolean	false	Include a base64-encoded crop of each detected face in the response. Forces JSON output.
padding	number (0…2)	0.15	Fraction by which to expand each crop's bounding box on every side (only used when multiple=true).
cropFormat	enum (jpeg \| png)	"jpeg"	Encoding for the per-face crops (jpeg / png).
cropQuality	number (1…100)	90	JPEG quality for the per-face crops (ignored for PNG).

Response

Modes: json. Cache: yes (24h TTL).