Skip to content

KeypointDetector

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:87

readonly inputSize: number

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:93

Width and height of the model input in pixels.


readonly landmarkNames: readonly ("nose" | "leftEye" | "rightEye" | "leftEar" | "rightEar" | "leftShoulder" | "rightShoulder" | "leftElbow" | "rightElbow" | "leftWrist" | "rightWrist" | "leftHip" | "rightHip" | "leftKnee" | "rightKnee" | "leftAnkle" | "rightAnkle")[]

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:89

The landmark names, in COCO_LANDMARKS order.


readonly schemas: object

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:97

Buffer layouts for encodeDetectKeypoints, at this runner’s size: the model input, and the list with room for maxDetections people. Pass one to root.createBuffer to make a buffer of the right type and size.

readonly input: Yolo26Input

readonly output: (maxDetections) => KeypointDetectionList

number

KeypointDetectionList


readonly skeleton: readonly readonly ["nose" | "leftEye" | "rightEye" | "leftEar" | "rightEar" | "leftShoulder" | "rightShoulder" | "leftElbow" | "rightElbow" | "leftWrist" | "rightWrist" | "leftHip" | "rightHip" | "leftKnee" | "rightKnee" | "leftAnkle" | "rightAnkle", "nose" | "leftEye" | "rightEye" | "leftEar" | "rightEar" | "leftShoulder" | "rightShoulder" | "leftElbow" | "rightElbow" | "leftWrist" | "rightWrist" | "leftHip" | "rightHip" | "leftKnee" | "rightKnee" | "leftAnkle" | "rightAnkle"][]

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:91

Pairs of landmark names to draw as limbs.

detectKeypoints(image, options?): Promise<KeypointDetection[]>

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:102

Image in, the people in it out, best first.

ImageBuffer

DetectKeypointsOptions

Promise<KeypointDetection[]>


dispose(): void

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:114

Frees the weights and every GPU buffer the runner holds.

void


encodeDetectKeypoints(options): void

Defined in: packages/runntime/src/zoo/tasks/cv/keypointDetection.ts:112

Records a detection from GPU memory into GPU memory, with no CPU copy. A texture source is fitted like an ImageBuffer; a buffer source is the model input as-is. The target gets the list of people, decoded on the GPU. Work submitted after it sees the result.

EncodeDetectKeypointsOptions

void

RunntimeError INVALID_ARGUMENT for a resource of the wrong format, usage or size, or an option out of range.