KeypointDetectorModel

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:23

What to load. Every field is optional; a missing modelPath comes from models.keypointDetection.YOLO26_POSE.DEFAULT.

Properties

inputSize?

readonly optional inputSize?: number

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:30

Input width and height in pixels, a multiple of 32. Default 640. Smaller runs faster and misses more small people.

---

modelPath?

readonly optional modelPath?: ModelPath

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:25

The weights: a safetensors export of a YOLO26 pose checkpoint.

---

resizeMode?

readonly optional resizeMode?: ResizeMode

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:32

How an image is fitted into the model input. Default letterbox.

---

variant?

readonly optional variant?: "x" | "m" | "n" | "s" | "l"

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:27

Model size, n to x. Default: read from the weights.