@svenflow/micro-handpose 0.1.0

package/dist/index.d.ts

@@ -0,0 +1,54 @@
+/** A 3D landmark point (x, y in [0,1] image coords, z is relative depth) */
+interface Landmark {
+    x: number;
+    y: number;
+    z: number;
+}
+/** Detection result for a single hand */
+interface HandposeResult {
+    /** Confidence score (0-1) that a hand is present */
+    score: number;
+    /** Whether this is a left or right hand */
+    handedness: 'left' | 'right';
+    /** 21 hand landmarks in order (wrist, thumb_cmc, ..., pinky_tip) */
+    landmarks: Landmark[];
+}
+/** Options for creating a handpose detector */
+interface HandposeOptions {
+    /** URL to fetch weights from. Defaults to bundled weights or CDN. */
+    weightsUrl?: string;
+    /** Minimum confidence score to return a detection (0-1). Default: 0.5 */
+    scoreThreshold?: number;
+}
+/** A handpose detector instance */
+interface Handpose {
+    /**
+     * Detect hand landmarks from an image source.
+     *
+     * Accepts: HTMLCanvasElement, OffscreenCanvas, ImageBitmap, HTMLImageElement,
+     * HTMLVideoElement, or ImageData.
+     *
+     * Returns null if no hand is detected (below score threshold).
+     */
+    detect: (source: HandposeInput) => Promise<HandposeResult | null>;
+    /** Dispose GPU resources */
+    dispose: () => void;
+}
+/** Accepted input types for detection */
+type HandposeInput = HTMLCanvasElement | OffscreenCanvas | ImageBitmap | HTMLImageElement | HTMLVideoElement | ImageData;
+
+/**
+ * Create a handpose detector.
+ *
+ * Downloads model weights and compiles the WebGPU pipeline.
+ * Call this once, then use `detect()` repeatedly.
+ *
+ * @example
+ * ```typescript
+ * const handpose = await createHandpose()
+ * const result = await handpose.detect(canvas)
+ * ```
+ */
+declare function createHandpose(options?: HandposeOptions): Promise<Handpose>;
+
+export { type Handpose, type HandposeOptions, type HandposeResult, type Landmark, createHandpose };