tempest-react-sdk 0.23.0 → 0.24.0

package/dist/vision/tasks/segmenter.js

@@ -1,189 +1,144 @@
-import { OrtSession as k } from "../core/session.js";
-import { loadImage as C } from "../io/image.js";
-import { resolveLabels as F } from "../labels.js";
-import { decodeYoloSeg as I } from "../postprocess/segmentation.js";
-import { letterbox as N, toFloat32 as U, toCHW as $, toFloat32Tensor as B } from "../preprocess/image.js";
-import { SegmentationResults as H, Boxes as L, Masks as W } from "../results.js";
-import { VisionTask as E } from "./base.js";
-import { RGBImage as M, Mask as D } from "../types.js";
-class v extends E {
-  constructor(n, s, o, t, e, r, i, a, _) {
-    super(n), this._head = s, this._labels = o, this._names = t, this._inputSize = e, this._confThreshold = r, this._iouThreshold = i, this._maxDetections = a, this._maskThreshold = _;
-  }
-  _head;
-  _labels;
-  _names;
-  _inputSize;
-  _confThreshold;
-  _iouThreshold;
-  _maxDetections;
-  _maskThreshold;
-  /** Load the model and resolve labels. */
-  static async create(n, s = {}) {
-    const o = s.head ?? "yolo-seg";
-    if (o !== "yolo-seg")
-      throw new Error(`Unsupported segmenter head '${o}'. Supported: 'yolo-seg'.`);
-    const t = await k.create(n, s), e = F(s.labels ?? "coco", {
-      numClasses: s.numClasses
-    }), r = {};
-    for (let i = 0; i < e.length; i++)
-      r[i] = e[i];
-    return new v(
-      t,
-      o,
-      e,
-      r,
-      s.inputSize ?? [640, 640],
-      s.confThreshold ?? 0.25,
-      s.iouThreshold ?? 0.45,
-      s.maxDetections ?? 300,
-      s.maskThreshold ?? 0.5
-    );
-  }
-  /** The decoder family used to interpret the model's output. */
-  get head() {
-    return this._head;
-  }
-  /** Class labels indexed by class id. */
-  get labels() {
-    return this._labels;
-  }
-  /** Class id → class name dict (matches Ultralytics' `model.names`). */
-  get names() {
-    return this._names;
-  }
-  /** Number of classes the model predicts. */
-  get numClasses() {
-    return this._labels.length;
-  }
-  /** Alias for {@link predict} (parity with PyTorch `nn.Module.__call__`). */
-  async call(n, s = {}) {
-    return this.predict(n, s);
-  }
-  /** Run instance segmentation on a single image. */
-  async predict(n, s = {}) {
-    const o = typeof n == "string" ? n : null, t = await C(n), { tensor: e, scale: r, padLeft: i, padTop: a } = this._preprocess(t), _ = await this._session.run({ [this._session.inputName]: e }), { perAnchor: f, prototypes: w } = this._splitOutputs(_), g = I(
-      f.data,
-      f.dims,
-      w.data,
-      w.dims,
-      {
-        numClasses: this._labels.length,
-        inputWidth: this._inputSize[0],
-        inputHeight: this._inputSize[1],
-        originalWidth: t.width,
-        originalHeight: t.height,
-        padLeft: i,
-        padTop: a,
-        scale: r,
-        confThreshold: s.confThreshold ?? this._confThreshold,
-        iouThreshold: s.iouThreshold ?? this._iouThreshold,
-        maxDetections: this._maxDetections,
-        maskThreshold: this._maskThreshold
-      }
-    ), d = (s.classes !== void 0 ? (() => {
-      const h = new Set(s.classes);
-      return g.filter((T) => h.has(T.classId));
-    })() : g).map(
-      (h) => this._buildResult(t, h.bbox, h.classId, h.confidence, h.mask)
-    ), m = [t.height, t.width];
-    return [
-      new H(
-        this._buildBoxes(d, m),
-        this._buildMasks(d, m),
-        d,
-        this._names,
-        t,
-        m,
-        o
-      )
-    ];
-  }
-  _preprocess(n) {
-    const [s, o] = this._inputSize, t = N(n, s, o), e = U(t.image), r = $(e, t.image.width, t.image.height, 3);
-    return {
-      tensor: B(r, [1, 3, t.image.height, t.image.width]),
-      scale: t.scale,
-      padLeft: t.padLeft,
-      padTop: t.padTop
-    };
-  }
-  _splitOutputs(n) {
-    let s, o;
-    for (const t of this._session.outputNames) {
-      const e = n[t];
-      e !== void 0 && (e.dims.length === 3 && s === void 0 ? s = e : e.dims.length === 4 && o === void 0 && (o = e));
-    }
-    if (s === void 0 || o === void 0) {
-      const t = this._session.outputNames.map(
-        (e) => `${e}: ${JSON.stringify(n[e]?.dims ?? [])}`
-      );
-      throw new Error(
-        `Segmenter expected one 3-D and one 4-D output, got [${t.join(", ")}].`
-      );
-    }
-    return { perAnchor: s, prototypes: o };
-  }
-  _buildResult(n, s, o, t, e) {
-    const [r, i, a, _] = s.asIntXyxy(), f = Math.max(0, r), w = Math.max(0, i), g = Math.min(n.width, a), y = Math.min(n.height, _);
-    let d, m = e;
-    if (g > f && y > w && e.data.length > 0) {
-      const T = g - f, O = y - w, l = Math.min(e.width, T), u = Math.min(e.height, O), b = new Uint8Array(l * u * 3);
-      for (let c = 0; c < u; c++) {
-        const p = ((w + c) * n.width + f) * 3, R = c * l * 3, z = c * e.width;
-        for (let x = 0; x < l; x++)
-          if (e.data[z + x] !== 0) {
-            const S = p + x * 3, A = R + x * 3;
-            b[A] = n.data[S], b[A + 1] = n.data[S + 1], b[A + 2] = n.data[S + 2];
-          }
-      }
-      if (d = new M(b, l, u), l !== e.width || u !== e.height) {
-        const c = new Uint8Array(l * u);
-        for (let p = 0; p < u; p++)
-          c.set(
-            e.data.subarray(p * e.width, p * e.width + l),
-            p * l
-          );
-        m = new D(c, l, u);
-      }
-    } else
-      m = new D(new Uint8Array(0), 0, 0), d = new M(new Uint8Array(0), 0, 0);
-    const h = this._names[o] ?? `class_${o}`;
-    return {
-      classId: o,
-      className: h,
-      confidence: t,
-      bbox: s,
-      cls: o,
-      name: h,
-      conf: t,
-      box: s,
-      mask: m,
-      segmentedImage: d
-    };
-  }
-  _buildBoxes(n, s) {
-    const o = n.length, t = new Float32Array(o * 4), e = new Int32Array(o), r = new Float32Array(o);
-    for (let i = 0; i < o; i++) {
-      const a = n[i];
-      t[i * 4] = a.bbox.x1, t[i * 4 + 1] = a.bbox.y1, t[i * 4 + 2] = a.bbox.x2, t[i * 4 + 3] = a.bbox.y2, e[i] = a.classId, r[i] = a.confidence;
-    }
-    return new L(t, e, r, s);
-  }
-  _buildMasks(n, s) {
-    const o = new Float32Array(n.length * 4);
-    for (let t = 0; t < n.length; t++) {
-      const e = n[t];
-      o[t * 4] = e.bbox.x1, o[t * 4 + 1] = e.bbox.y1, o[t * 4 + 2] = e.bbox.x2, o[t * 4 + 3] = e.bbox.y2;
-    }
-    return new W(
-      n.map((t) => t.mask),
-      o,
-      s
-    );
-  }
-}
-export {
-  v as Segmenter
+import { Mask as e, RGBImage as t } from "../types.js";
+import { Boxes as n, Masks as r, SegmentationResults as i } from "../results.js";
+import { resolveLabels as a } from "../labels.js";
+import { OrtSession as o } from "../core/session.js";
+import { loadImage as s } from "../io/image.js";
+import { letterbox as c, toCHW as l, toFloat32 as u, toFloat32Tensor as d } from "../preprocess/image.js";
+import { decodeYoloSeg as f } from "../postprocess/segmentation.js";
+import { VisionTask as p } from "./base.js";
+//#region src/vision/tasks/segmenter.ts
+var m = class m extends p {
+	_head;
+	_labels;
+	_names;
+	_inputSize;
+	_confThreshold;
+	_iouThreshold;
+	_maxDetections;
+	_maskThreshold;
+	constructor(e, t, n, r, i, a, o, s, c) {
+		super(e), this._head = t, this._labels = n, this._names = r, this._inputSize = i, this._confThreshold = a, this._iouThreshold = o, this._maxDetections = s, this._maskThreshold = c;
+	}
+	static async create(e, t = {}) {
+		let n = t.head ?? "yolo-seg";
+		if (n !== "yolo-seg") throw Error(`Unsupported segmenter head '${n}'. Supported: 'yolo-seg'.`);
+		let r = await o.create(e, t), i = a(t.labels ?? "coco", { numClasses: t.numClasses }), s = {};
+		for (let e = 0; e < i.length; e++) s[e] = i[e];
+		return new m(r, n, i, s, t.inputSize ?? [640, 640], t.confThreshold ?? .25, t.iouThreshold ?? .45, t.maxDetections ?? 300, t.maskThreshold ?? .5);
+	}
+	get head() {
+		return this._head;
+	}
+	get labels() {
+		return this._labels;
+	}
+	get names() {
+		return this._names;
+	}
+	get numClasses() {
+		return this._labels.length;
+	}
+	async call(e, t = {}) {
+		return this.predict(e, t);
+	}
+	async predict(e, t = {}) {
+		let n = typeof e == "string" ? e : null, r = await s(e), { tensor: a, scale: o, padLeft: c, padTop: l } = this._preprocess(r), u = await this._session.run({ [this._session.inputName]: a }), { perAnchor: d, prototypes: p } = this._splitOutputs(u), m = f(d.data, d.dims, p.data, p.dims, {
+			numClasses: this._labels.length,
+			inputWidth: this._inputSize[0],
+			inputHeight: this._inputSize[1],
+			originalWidth: r.width,
+			originalHeight: r.height,
+			padLeft: c,
+			padTop: l,
+			scale: o,
+			confThreshold: t.confThreshold ?? this._confThreshold,
+			iouThreshold: t.iouThreshold ?? this._iouThreshold,
+			maxDetections: this._maxDetections,
+			maskThreshold: this._maskThreshold
+		}), h = (t.classes === void 0 ? m : (() => {
+			let e = new Set(t.classes);
+			return m.filter((t) => e.has(t.classId));
+		})()).map((e) => this._buildResult(r, e.bbox, e.classId, e.confidence, e.mask)), g = [r.height, r.width];
+		return [new i(this._buildBoxes(h, g), this._buildMasks(h, g), h, this._names, r, g, n)];
+	}
+	_preprocess(e) {
+		let [t, n] = this._inputSize, r = c(e, t, n);
+		return {
+			tensor: d(l(u(r.image), r.image.width, r.image.height, 3), [
+				1,
+				3,
+				r.image.height,
+				r.image.width
+			]),
+			scale: r.scale,
+			padLeft: r.padLeft,
+			padTop: r.padTop
+		};
+	}
+	_splitOutputs(e) {
+		let t, n;
+		for (let r of this._session.outputNames) {
+			let i = e[r];
+			i !== void 0 && (i.dims.length === 3 && t === void 0 ? t = i : i.dims.length === 4 && n === void 0 && (n = i));
+		}
+		if (t === void 0 || n === void 0) {
+			let t = this._session.outputNames.map((t) => `${t}: ${JSON.stringify(e[t]?.dims ?? [])}`);
+			throw Error(`Segmenter expected one 3-D and one 4-D output, got [${t.join(", ")}].`);
+		}
+		return {
+			perAnchor: t,
+			prototypes: n
+		};
+	}
+	_buildResult(n, r, i, a, o) {
+		let [s, c, l, u] = r.asIntXyxy(), d = Math.max(0, s), f = Math.max(0, c), p = Math.min(n.width, l), m = Math.min(n.height, u), h, g = o;
+		if (p > d && m > f && o.data.length > 0) {
+			let r = p - d, i = m - f, a = Math.min(o.width, r), s = Math.min(o.height, i), c = new Uint8Array(a * s * 3);
+			for (let e = 0; e < s; e++) {
+				let t = ((f + e) * n.width + d) * 3, r = e * a * 3, i = e * o.width;
+				for (let e = 0; e < a; e++) if (o.data[i + e] !== 0) {
+					let i = t + e * 3, a = r + e * 3;
+					c[a] = n.data[i], c[a + 1] = n.data[i + 1], c[a + 2] = n.data[i + 2];
+				}
+			}
+			if (h = new t(c, a, s), a !== o.width || s !== o.height) {
+				let t = new Uint8Array(a * s);
+				for (let e = 0; e < s; e++) t.set(o.data.subarray(e * o.width, e * o.width + a), e * a);
+				g = new e(t, a, s);
+			}
+		} else g = new e(/* @__PURE__ */ new Uint8Array(), 0, 0), h = new t(/* @__PURE__ */ new Uint8Array(), 0, 0);
+		let _ = this._names[i] ?? `class_${i}`;
+		return {
+			classId: i,
+			className: _,
+			confidence: a,
+			bbox: r,
+			cls: i,
+			name: _,
+			conf: a,
+			box: r,
+			mask: g,
+			segmentedImage: h
+		};
+	}
+	_buildBoxes(e, t) {
+		let r = e.length, i = new Float32Array(r * 4), a = new Int32Array(r), o = new Float32Array(r);
+		for (let t = 0; t < r; t++) {
+			let n = e[t];
+			i[t * 4] = n.bbox.x1, i[t * 4 + 1] = n.bbox.y1, i[t * 4 + 2] = n.bbox.x2, i[t * 4 + 3] = n.bbox.y2, a[t] = n.classId, o[t] = n.confidence;
+		}
+		return new n(i, a, o, t);
+	}
+	_buildMasks(e, t) {
+		let n = new Float32Array(e.length * 4);
+		for (let t = 0; t < e.length; t++) {
+			let r = e[t];
+			n[t * 4] = r.bbox.x1, n[t * 4 + 1] = r.bbox.y1, n[t * 4 + 2] = r.bbox.x2, n[t * 4 + 3] = r.bbox.y2;
+		}
+		return new r(e.map((e) => e.mask), n, t);
+	}
 };
-//# sourceMappingURL=segmenter.js.map
+//#endregion
+export { m as Segmenter };
+
+//# sourceMappingURL=segmenter.js.map

package/dist/vision/tasks/segmenter.js.map

@@ -1 +1 @@
-{"version":3,"file":"segmenter.js","sources":["../../../src/vision/tasks/segmenter.ts"],"sourcesContent":["/**\n * Instance-segmentation task using YOLO seg ONNX models (v8-seg / v11-seg / ...).\n */\n\nimport type * as ort from \"onnxruntime-web\";\n\nimport { type ModelSource, type OrtSessionOptions, OrtSession } from \"../core/session\";\nimport { type ImageInput, loadImage } from \"../io/image\";\nimport { type LabelSpec, resolveLabels } from \"../labels\";\nimport { decodeYoloSeg } from \"../postprocess/segmentation\";\nimport { letterbox, toCHW, toFloat32, toFloat32Tensor } from \"../preprocess/image\";\nimport { Boxes, Masks, SegmentationResults } from \"../results\";\nimport { VisionTask } from \"./base\";\nimport { type BoundingBox, type SegmentationResult, Mask, RGBImage } from \"../types\";\n\n/**\n * Decoder family for the segmentation head.\n *\n * - `\"yolo-seg\"`: YOLO instance-segmentation head with two outputs —\n *   `[1, 4 + nc + nm, N]` per-anchor predictions plus `[1, nm, mh, mw]`\n *   prototype masks. Covers YOLOv8-seg, v11-seg, v26-seg.\n *\n * The SDK does **not** auto-detect this — the caller is responsible for\n * picking a head that matches their export.\n */\nexport type SegmenterHead = \"yolo-seg\";\n\nexport interface SegmenterOptions extends OrtSessionOptions {\n    /**\n     * Decoder family for the segmentation head. Default `\"yolo-seg\"` covers\n     * YOLOv8-seg/v11-seg/v26-seg.\n     */\n    readonly head?: SegmenterHead;\n    /** Class label spec — see {@link resolveLabels}. Defaults to the COCO 80-class preset. */\n    readonly labels?: LabelSpec;\n    /** Number of classes — used to validate the supplied labels. */\n    readonly numClasses?: number;\n    /** Model input `[width, height]` for letterboxing. Defaults to `[640, 640]`. */\n    readonly inputSize?: readonly [number, number];\n    /** Default minimum class score to keep a candidate. */\n    readonly confThreshold?: number;\n    /** Default IoU threshold for non-maximum suppression. */\n    readonly iouThreshold?: number;\n    /** Maximum number of instances per image. */\n    readonly maxDetections?: number;\n    /** Probability cutoff applied to soft masks. Defaults to `0.5`. */\n    readonly maskThreshold?: number;\n}\n\nexport interface SegmenterPredictOptions {\n    readonly confThreshold?: number;\n    readonly iouThreshold?: number;\n    /**\n     * If set, keep only instances whose `classId` is in this list.\n     * Mirrors Ultralytics' `model.predict(img, classes=[0, 16])`.\n     */\n    readonly classes?: readonly number[];\n}\n\n/**\n * Instance segmenter for YOLO seg ONNX models (v8-seg / v11-seg / ...).\n *\n * The model is expected to expose two outputs:\n *\n * 1. `output0`: `(1, 4 + numClasses + numMaskCoefs, numAnchors)` — per-anchor\n *    predictions (boxes, class scores, mask coefficients).\n * 2. `output1`: `(1, numMaskCoefs, maskH, maskW)` — prototype masks.\n *\n * `predict()` returns `Promise<SegmentationResults[]>` (length 1 for a\n * single image), mirroring Ultralytics' API. The envelope exposes:\n *\n * - `boxes`: bulk numpy view (`xyxy`, `xywh`, `xyxyn`, `xywhn`, `cls`, `conf`).\n * - `masks`: per-instance binary masks cropped to each box.\n * - per-instance {@link SegmentationResult} via iteration.\n *\n * @example\n * ```typescript\n * const seg = await Segmenter.create(\"/models/yolov8n-seg.onnx\");\n * const r = (await seg.predict(\"/images/street.jpg\"))[0];\n * for (const inst of r) {\n *   console.log(inst.cls, inst.conf, inst.box.xyxy);\n * }\n * ```\n */\nexport class Segmenter extends VisionTask {\n    private constructor(\n        session: OrtSession,\n        private readonly _head: SegmenterHead,\n        private readonly _labels: readonly string[],\n        private readonly _names: Readonly<Record<number, string>>,\n        private readonly _inputSize: readonly [number, number],\n        private readonly _confThreshold: number,\n        private readonly _iouThreshold: number,\n        private readonly _maxDetections: number,\n        private readonly _maskThreshold: number,\n    ) {\n        super(session);\n    }\n\n    /** Load the model and resolve labels. */\n    static async create(model: ModelSource, options: SegmenterOptions = {}): Promise<Segmenter> {\n        const head: SegmenterHead = options.head ?? \"yolo-seg\";\n        if (head !== \"yolo-seg\") {\n            throw new Error(`Unsupported segmenter head '${head}'. Supported: 'yolo-seg'.`);\n        }\n        const session = await OrtSession.create(model, options);\n        const labels = resolveLabels(options.labels ?? \"coco\", {\n            numClasses: options.numClasses,\n        });\n        const names: Record<number, string> = {};\n        for (let i = 0; i < labels.length; i++) {\n            names[i] = labels[i] as string;\n        }\n        return new Segmenter(\n            session,\n            head,\n            labels,\n            names,\n            options.inputSize ?? [640, 640],\n            options.confThreshold ?? 0.25,\n            options.iouThreshold ?? 0.45,\n            options.maxDetections ?? 300,\n            options.maskThreshold ?? 0.5,\n        );\n    }\n\n    /** The decoder family used to interpret the model's output. */\n    get head(): SegmenterHead {\n        return this._head;\n    }\n\n    /** Class labels indexed by class id. */\n    get labels(): readonly string[] {\n        return this._labels;\n    }\n\n    /** Class id → class name dict (matches Ultralytics' `model.names`). */\n    get names(): Readonly<Record<number, string>> {\n        return this._names;\n    }\n\n    /** Number of classes the model predicts. */\n    get numClasses(): number {\n        return this._labels.length;\n    }\n\n    /** Alias for {@link predict} (parity with PyTorch `nn.Module.__call__`). */\n    async call(\n        image: ImageInput,\n        options: SegmenterPredictOptions = {},\n    ): Promise<SegmentationResults[]> {\n        return this.predict(image, options);\n    }\n\n    /** Run instance segmentation on a single image. */\n    async predict(\n        image: ImageInput,\n        options: SegmenterPredictOptions = {},\n    ): Promise<SegmentationResults[]> {\n        const path = typeof image === \"string\" ? image : null;\n        const original = await loadImage(image);\n        const { tensor, scale, padLeft, padTop } = this._preprocess(original);\n        const outputs = await this._session.run({ [this._session.inputName]: tensor });\n\n        const { perAnchor, prototypes } = this._splitOutputs(outputs);\n\n        const decodedAll = decodeYoloSeg(\n            perAnchor.data as Float32Array,\n            perAnchor.dims,\n            prototypes.data as Float32Array,\n            prototypes.dims,\n            {\n                numClasses: this._labels.length,\n                inputWidth: this._inputSize[0],\n                inputHeight: this._inputSize[1],\n                originalWidth: original.width,\n                originalHeight: original.height,\n                padLeft,\n                padTop,\n                scale,\n                confThreshold: options.confThreshold ?? this._confThreshold,\n                iouThreshold: options.iouThreshold ?? this._iouThreshold,\n                maxDetections: this._maxDetections,\n                maskThreshold: this._maskThreshold,\n            },\n        );\n\n        const decoded =\n            options.classes !== undefined\n                ? (() => {\n                      const allowed = new Set(options.classes);\n                      return decodedAll.filter((d) => allowed.has(d.classId));\n                  })()\n                : decodedAll;\n\n        const detections = decoded.map((d) =>\n            this._buildResult(original, d.bbox, d.classId, d.confidence, d.mask),\n        );\n\n        const orig: readonly [number, number] = [original.height, original.width];\n        return [\n            new SegmentationResults(\n                this._buildBoxes(detections, orig),\n                this._buildMasks(detections, orig),\n                detections,\n                this._names,\n                original,\n                orig,\n                path,\n            ),\n        ];\n    }\n\n    private _preprocess(image: RGBImage): {\n        tensor: ort.Tensor;\n        scale: number;\n        padLeft: number;\n        padTop: number;\n    } {\n        const [tw, th] = this._inputSize;\n        const lb = letterbox(image, tw, th);\n        const f32 = toFloat32(lb.image);\n        const chw = toCHW(f32, lb.image.width, lb.image.height, 3);\n        return {\n            tensor: toFloat32Tensor(chw, [1, 3, lb.image.height, lb.image.width]),\n            scale: lb.scale,\n            padLeft: lb.padLeft,\n            padTop: lb.padTop,\n        };\n    }\n\n    private _splitOutputs(outputs: Record<string, ort.Tensor>): {\n        perAnchor: ort.Tensor;\n        prototypes: ort.Tensor;\n    } {\n        let perAnchor: ort.Tensor | undefined;\n        let prototypes: ort.Tensor | undefined;\n        for (const name of this._session.outputNames) {\n            const t = outputs[name];\n            if (t === undefined) continue;\n            if (t.dims.length === 3 && perAnchor === undefined) {\n                perAnchor = t;\n            } else if (t.dims.length === 4 && prototypes === undefined) {\n                prototypes = t;\n            }\n        }\n        if (perAnchor === undefined || prototypes === undefined) {\n            const shapes = this._session.outputNames.map(\n                (n) => `${n}: ${JSON.stringify(outputs[n]?.dims ?? [])}`,\n            );\n            throw new Error(\n                `Segmenter expected one 3-D and one 4-D output, got [${shapes.join(\", \")}].`,\n            );\n        }\n        return { perAnchor, prototypes };\n    }\n\n    private _buildResult(\n        original: RGBImage,\n        bbox: BoundingBox,\n        classId: number,\n        confidence: number,\n        mask: Mask,\n    ): SegmentationResult {\n        const [x1, y1, x2, y2] = bbox.asIntXyxy();\n        const cx1 = Math.max(0, x1);\n        const cy1 = Math.max(0, y1);\n        const cx2 = Math.min(original.width, x2);\n        const cy2 = Math.min(original.height, y2);\n\n        let segmentedImage: RGBImage;\n        let finalMask = mask;\n        if (cx2 > cx1 && cy2 > cy1 && mask.data.length > 0) {\n            const cropW = cx2 - cx1;\n            const cropH = cy2 - cy1;\n            const mw = Math.min(mask.width, cropW);\n            const mh = Math.min(mask.height, cropH);\n            const segData = new Uint8Array(mw * mh * 3);\n            for (let row = 0; row < mh; row++) {\n                const srcRowOffset = ((cy1 + row) * original.width + cx1) * 3;\n                const dstRowOffset = row * mw * 3;\n                const maskRowOffset = row * mask.width;\n                for (let col = 0; col < mw; col++) {\n                    const m = mask.data[maskRowOffset + col];\n                    if (m !== 0) {\n                        const s = srcRowOffset + col * 3;\n                        const d = dstRowOffset + col * 3;\n                        segData[d] = original.data[s];\n                        segData[d + 1] = original.data[s + 1];\n                        segData[d + 2] = original.data[s + 2];\n                    }\n                }\n            }\n            segmentedImage = new RGBImage(segData, mw, mh);\n            if (mw !== mask.width || mh !== mask.height) {\n                const trimmed = new Uint8Array(mw * mh);\n                for (let row = 0; row < mh; row++) {\n                    trimmed.set(\n                        mask.data.subarray(row * mask.width, row * mask.width + mw),\n                        row * mw,\n                    );\n                }\n                finalMask = new Mask(trimmed, mw, mh);\n            }\n        } else {\n            finalMask = new Mask(new Uint8Array(0), 0, 0);\n            segmentedImage = new RGBImage(new Uint8Array(0), 0, 0);\n        }\n\n        const className = this._names[classId] ?? `class_${classId}`;\n\n        return {\n            classId,\n            className,\n            confidence,\n            bbox,\n            cls: classId,\n            name: className,\n            conf: confidence,\n            box: bbox,\n            mask: finalMask,\n            segmentedImage,\n        };\n    }\n\n    private _buildBoxes(\n        detections: readonly SegmentationResult[],\n        origShape: readonly [number, number],\n    ): Boxes {\n        const n = detections.length;\n        const xyxy = new Float32Array(n * 4);\n        const cls = new Int32Array(n);\n        const conf = new Float32Array(n);\n        for (let i = 0; i < n; i++) {\n            const d = detections[i] as SegmentationResult;\n            xyxy[i * 4] = d.bbox.x1;\n            xyxy[i * 4 + 1] = d.bbox.y1;\n            xyxy[i * 4 + 2] = d.bbox.x2;\n            xyxy[i * 4 + 3] = d.bbox.y2;\n            cls[i] = d.classId;\n            conf[i] = d.confidence;\n        }\n        return new Boxes(xyxy, cls, conf, origShape);\n    }\n\n    private _buildMasks(\n        detections: readonly SegmentationResult[],\n        origShape: readonly [number, number],\n    ): Masks {\n        const xyxy = new Float32Array(detections.length * 4);\n        for (let i = 0; i < detections.length; i++) {\n            const d = detections[i] as SegmentationResult;\n            xyxy[i * 4] = d.bbox.x1;\n            xyxy[i * 4 + 1] = d.bbox.y1;\n            xyxy[i * 4 + 2] = d.bbox.x2;\n            xyxy[i * 4 + 3] = d.bbox.y2;\n        }\n        return new Masks(\n            detections.map((d) => d.mask),\n            xyxy,\n            origShape,\n        );\n    }\n}\n"],"names":["Segmenter","VisionTask","session","_head","_labels","_names","_inputSize","_confThreshold","_iouThreshold","_maxDetections","_maskThreshold","model","options","head","OrtSession","labels","resolveLabels","names","image","path","original","loadImage","tensor","scale","padLeft","padTop","outputs","perAnchor","prototypes","decodedAll","decodeYoloSeg","detections","allowed","d","orig","SegmentationResults","tw","th","lb","letterbox","f32","toFloat32","chw","toCHW","toFloat32Tensor","name","t","shapes","n","bbox","classId","confidence","mask","x1","y1","x2","y2","cx1","cy1","cx2","cy2","segmentedImage","finalMask","cropW","cropH","mw","mh","segData","row","srcRowOffset","dstRowOffset","maskRowOffset","col","s","RGBImage","trimmed","Mask","className","origShape","xyxy","cls","conf","Boxes","i","Masks"],"mappings":";;;;;;;;AAoFO,MAAMA,UAAkBC,EAAW;AAAA,EAC9B,YACJC,GACiBC,GACAC,GACAC,GACAC,GACAC,GACAC,GACAC,GACAC,GACnB;AACE,UAAMR,CAAO,GATI,KAAA,QAAAC,GACA,KAAA,UAAAC,GACA,KAAA,SAAAC,GACA,KAAA,aAAAC,GACA,KAAA,iBAAAC,GACA,KAAA,gBAAAC,GACA,KAAA,iBAAAC,GACA,KAAA,iBAAAC;AAAA,EAGrB;AAAA,EAVqB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAMrB,aAAa,OAAOC,GAAoBC,IAA4B,IAAwB;AACxF,UAAMC,IAAsBD,EAAQ,QAAQ;AAC5C,QAAIC,MAAS;AACT,YAAM,IAAI,MAAM,+BAA+BA,CAAI,2BAA2B;AAElF,UAAMX,IAAU,MAAMY,EAAW,OAAOH,GAAOC,CAAO,GAChDG,IAASC,EAAcJ,EAAQ,UAAU,QAAQ;AAAA,MACnD,YAAYA,EAAQ;AAAA,IAAA,CACvB,GACKK,IAAgC,CAAA;AACtC,aAAS,IAAI,GAAG,IAAIF,EAAO,QAAQ;AAC/B,MAAAE,EAAM,CAAC,IAAIF,EAAO,CAAC;AAEvB,WAAO,IAAIf;AAAA,MACPE;AAAA,MACAW;AAAA,MACAE;AAAA,MACAE;AAAA,MACAL,EAAQ,aAAa,CAAC,KAAK,GAAG;AAAA,MAC9BA,EAAQ,iBAAiB;AAAA,MACzBA,EAAQ,gBAAgB;AAAA,MACxBA,EAAQ,iBAAiB;AAAA,MACzBA,EAAQ,iBAAiB;AAAA,IAAA;AAAA,EAEjC;AAAA;AAAA,EAGA,IAAI,OAAsB;AACtB,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA,EAGA,IAAI,SAA4B;AAC5B,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA,EAGA,IAAI,QAA0C;AAC1C,WAAO,KAAK;AAAA,EAChB;AAAA;AAAA,EAGA,IAAI,aAAqB;AACrB,WAAO,KAAK,QAAQ;AAAA,EACxB;AAAA;AAAA,EAGA,MAAM,KACFM,GACAN,IAAmC,IACL;AAC9B,WAAO,KAAK,QAAQM,GAAON,CAAO;AAAA,EACtC;AAAA;AAAA,EAGA,MAAM,QACFM,GACAN,IAAmC,IACL;AAC9B,UAAMO,IAAO,OAAOD,KAAU,WAAWA,IAAQ,MAC3CE,IAAW,MAAMC,EAAUH,CAAK,GAChC,EAAE,QAAAI,GAAQ,OAAAC,GAAO,SAAAC,GAAS,QAAAC,MAAW,KAAK,YAAYL,CAAQ,GAC9DM,IAAU,MAAM,KAAK,SAAS,IAAI,EAAE,CAAC,KAAK,SAAS,SAAS,GAAGJ,GAAQ,GAEvE,EAAE,WAAAK,GAAW,YAAAC,EAAA,IAAe,KAAK,cAAcF,CAAO,GAEtDG,IAAaC;AAAA,MACfH,EAAU;AAAA,MACVA,EAAU;AAAA,MACVC,EAAW;AAAA,MACXA,EAAW;AAAA,MACX;AAAA,QACI,YAAY,KAAK,QAAQ;AAAA,QACzB,YAAY,KAAK,WAAW,CAAC;AAAA,QAC7B,aAAa,KAAK,WAAW,CAAC;AAAA,QAC9B,eAAeR,EAAS;AAAA,QACxB,gBAAgBA,EAAS;AAAA,QACzB,SAAAI;AAAA,QACA,QAAAC;AAAA,QACA,OAAAF;AAAA,QACA,eAAeX,EAAQ,iBAAiB,KAAK;AAAA,QAC7C,cAAcA,EAAQ,gBAAgB,KAAK;AAAA,QAC3C,eAAe,KAAK;AAAA,QACpB,eAAe,KAAK;AAAA,MAAA;AAAA,IACxB,GAWEmB,KAPFnB,EAAQ,YAAY,UACb,MAAM;AACH,YAAMoB,IAAU,IAAI,IAAIpB,EAAQ,OAAO;AACvC,aAAOiB,EAAW,OAAO,CAACI,MAAMD,EAAQ,IAAIC,EAAE,OAAO,CAAC;AAAA,IAC1D,OACAJ,GAEiB;AAAA,MAAI,CAACI,MAC5B,KAAK,aAAab,GAAUa,EAAE,MAAMA,EAAE,SAASA,EAAE,YAAYA,EAAE,IAAI;AAAA,IAAA,GAGjEC,IAAkC,CAACd,EAAS,QAAQA,EAAS,KAAK;AACxE,WAAO;AAAA,MACH,IAAIe;AAAA,QACA,KAAK,YAAYJ,GAAYG,CAAI;AAAA,QACjC,KAAK,YAAYH,GAAYG,CAAI;AAAA,QACjCH;AAAA,QACA,KAAK;AAAA,QACLX;AAAA,QACAc;AAAA,QACAf;AAAA,MAAA;AAAA,IACJ;AAAA,EAER;AAAA,EAEQ,YAAYD,GAKlB;AACE,UAAM,CAACkB,GAAIC,CAAE,IAAI,KAAK,YAChBC,IAAKC,EAAUrB,GAAOkB,GAAIC,CAAE,GAC5BG,IAAMC,EAAUH,EAAG,KAAK,GACxBI,IAAMC,EAAMH,GAAKF,EAAG,MAAM,OAAOA,EAAG,MAAM,QAAQ,CAAC;AACzD,WAAO;AAAA,MACH,QAAQM,EAAgBF,GAAK,CAAC,GAAG,GAAGJ,EAAG,MAAM,QAAQA,EAAG,MAAM,KAAK,CAAC;AAAA,MACpE,OAAOA,EAAG;AAAA,MACV,SAASA,EAAG;AAAA,MACZ,QAAQA,EAAG;AAAA,IAAA;AAAA,EAEnB;AAAA,EAEQ,cAAcZ,GAGpB;AACE,QAAIC,GACAC;AACJ,eAAWiB,KAAQ,KAAK,SAAS,aAAa;AAC1C,YAAMC,IAAIpB,EAAQmB,CAAI;AACtB,MAAIC,MAAM,WACNA,EAAE,KAAK,WAAW,KAAKnB,MAAc,SACrCA,IAAYmB,IACLA,EAAE,KAAK,WAAW,KAAKlB,MAAe,WAC7CA,IAAakB;AAAA,IAErB;AACA,QAAInB,MAAc,UAAaC,MAAe,QAAW;AACrD,YAAMmB,IAAS,KAAK,SAAS,YAAY;AAAA,QACrC,CAACC,MAAM,GAAGA,CAAC,KAAK,KAAK,UAAUtB,EAAQsB,CAAC,GAAG,QAAQ,CAAA,CAAE,CAAC;AAAA,MAAA;AAE1D,YAAM,IAAI;AAAA,QACN,uDAAuDD,EAAO,KAAK,IAAI,CAAC;AAAA,MAAA;AAAA,IAEhF;AACA,WAAO,EAAE,WAAApB,GAAW,YAAAC,EAAA;AAAA,EACxB;AAAA,EAEQ,aACJR,GACA6B,GACAC,GACAC,GACAC,GACkB;AAClB,UAAM,CAACC,GAAIC,GAAIC,GAAIC,CAAE,IAAIP,EAAK,UAAA,GACxBQ,IAAM,KAAK,IAAI,GAAGJ,CAAE,GACpBK,IAAM,KAAK,IAAI,GAAGJ,CAAE,GACpBK,IAAM,KAAK,IAAIvC,EAAS,OAAOmC,CAAE,GACjCK,IAAM,KAAK,IAAIxC,EAAS,QAAQoC,CAAE;AAExC,QAAIK,GACAC,IAAYV;AAChB,QAAIO,IAAMF,KAAOG,IAAMF,KAAON,EAAK,KAAK,SAAS,GAAG;AAChD,YAAMW,IAAQJ,IAAMF,GACdO,IAAQJ,IAAMF,GACdO,IAAK,KAAK,IAAIb,EAAK,OAAOW,CAAK,GAC/BG,IAAK,KAAK,IAAId,EAAK,QAAQY,CAAK,GAChCG,IAAU,IAAI,WAAWF,IAAKC,IAAK,CAAC;AAC1C,eAASE,IAAM,GAAGA,IAAMF,GAAIE,KAAO;AAC/B,cAAMC,MAAiBX,IAAMU,KAAOhD,EAAS,QAAQqC,KAAO,GACtDa,IAAeF,IAAMH,IAAK,GAC1BM,IAAgBH,IAAMhB,EAAK;AACjC,iBAASoB,IAAM,GAAGA,IAAMP,GAAIO;AAExB,cADUpB,EAAK,KAAKmB,IAAgBC,CAAG,MAC7B,GAAG;AACT,kBAAMC,IAAIJ,IAAeG,IAAM,GACzBvC,IAAIqC,IAAeE,IAAM;AAC/B,YAAAL,EAAQlC,CAAC,IAAIb,EAAS,KAAKqD,CAAC,GAC5BN,EAAQlC,IAAI,CAAC,IAAIb,EAAS,KAAKqD,IAAI,CAAC,GACpCN,EAAQlC,IAAI,CAAC,IAAIb,EAAS,KAAKqD,IAAI,CAAC;AAAA,UACxC;AAAA,MAER;AAEA,UADAZ,IAAiB,IAAIa,EAASP,GAASF,GAAIC,CAAE,GACzCD,MAAOb,EAAK,SAASc,MAAOd,EAAK,QAAQ;AACzC,cAAMuB,IAAU,IAAI,WAAWV,IAAKC,CAAE;AACtC,iBAASE,IAAM,GAAGA,IAAMF,GAAIE;AACxB,UAAAO,EAAQ;AAAA,YACJvB,EAAK,KAAK,SAASgB,IAAMhB,EAAK,OAAOgB,IAAMhB,EAAK,QAAQa,CAAE;AAAA,YAC1DG,IAAMH;AAAA,UAAA;AAGd,QAAAH,IAAY,IAAIc,EAAKD,GAASV,GAAIC,CAAE;AAAA,MACxC;AAAA,IACJ;AACI,MAAAJ,IAAY,IAAIc,EAAK,IAAI,WAAW,CAAC,GAAG,GAAG,CAAC,GAC5Cf,IAAiB,IAAIa,EAAS,IAAI,WAAW,CAAC,GAAG,GAAG,CAAC;AAGzD,UAAMG,IAAY,KAAK,OAAO3B,CAAO,KAAK,SAASA,CAAO;AAE1D,WAAO;AAAA,MACH,SAAAA;AAAA,MACA,WAAA2B;AAAA,MACA,YAAA1B;AAAA,MACA,MAAAF;AAAA,MACA,KAAKC;AAAA,MACL,MAAM2B;AAAA,MACN,MAAM1B;AAAA,MACN,KAAKF;AAAA,MACL,MAAMa;AAAA,MACN,gBAAAD;AAAA,IAAA;AAAA,EAER;AAAA,EAEQ,YACJ9B,GACA+C,GACK;AACL,UAAM9B,IAAIjB,EAAW,QACfgD,IAAO,IAAI,aAAa/B,IAAI,CAAC,GAC7BgC,IAAM,IAAI,WAAWhC,CAAC,GACtBiC,IAAO,IAAI,aAAajC,CAAC;AAC/B,aAAS,IAAI,GAAG,IAAIA,GAAG,KAAK;AACxB,YAAMf,IAAIF,EAAW,CAAC;AACtB,MAAAgD,EAAK,IAAI,CAAC,IAAI9C,EAAE,KAAK,IACrB8C,EAAK,IAAI,IAAI,CAAC,IAAI9C,EAAE,KAAK,IACzB8C,EAAK,IAAI,IAAI,CAAC,IAAI9C,EAAE,KAAK,IACzB8C,EAAK,IAAI,IAAI,CAAC,IAAI9C,EAAE,KAAK,IACzB+C,EAAI,CAAC,IAAI/C,EAAE,SACXgD,EAAK,CAAC,IAAIhD,EAAE;AAAA,IAChB;AACA,WAAO,IAAIiD,EAAMH,GAAMC,GAAKC,GAAMH,CAAS;AAAA,EAC/C;AAAA,EAEQ,YACJ/C,GACA+C,GACK;AACL,UAAMC,IAAO,IAAI,aAAahD,EAAW,SAAS,CAAC;AACnD,aAASoD,IAAI,GAAGA,IAAIpD,EAAW,QAAQoD,KAAK;AACxC,YAAMlD,IAAIF,EAAWoD,CAAC;AACtB,MAAAJ,EAAKI,IAAI,CAAC,IAAIlD,EAAE,KAAK,IACrB8C,EAAKI,IAAI,IAAI,CAAC,IAAIlD,EAAE,KAAK,IACzB8C,EAAKI,IAAI,IAAI,CAAC,IAAIlD,EAAE,KAAK,IACzB8C,EAAKI,IAAI,IAAI,CAAC,IAAIlD,EAAE,KAAK;AAAA,IAC7B;AACA,WAAO,IAAImD;AAAA,MACPrD,EAAW,IAAI,CAACE,MAAMA,EAAE,IAAI;AAAA,MAC5B8C;AAAA,MACAD;AAAA,IAAA;AAAA,EAER;AACJ;"}
+{"version":3,"file":"segmenter.js","names":[],"sources":["../../../src/vision/tasks/segmenter.ts"],"sourcesContent":["/**\n * Instance-segmentation task using YOLO seg ONNX models (v8-seg / v11-seg / ...).\n */\n\nimport type * as ort from \"onnxruntime-web\";\n\nimport { type ModelSource, type OrtSessionOptions, OrtSession } from \"../core/session\";\nimport { type ImageInput, loadImage } from \"../io/image\";\nimport { type LabelSpec, resolveLabels } from \"../labels\";\nimport { decodeYoloSeg } from \"../postprocess/segmentation\";\nimport { letterbox, toCHW, toFloat32, toFloat32Tensor } from \"../preprocess/image\";\nimport { Boxes, Masks, SegmentationResults } from \"../results\";\nimport { VisionTask } from \"./base\";\nimport { type BoundingBox, type SegmentationResult, Mask, RGBImage } from \"../types\";\n\n/**\n * Decoder family for the segmentation head.\n *\n * - `\"yolo-seg\"`: YOLO instance-segmentation head with two outputs —\n *   `[1, 4 + nc + nm, N]` per-anchor predictions plus `[1, nm, mh, mw]`\n *   prototype masks. Covers YOLOv8-seg, v11-seg, v26-seg.\n *\n * The SDK does **not** auto-detect this — the caller is responsible for\n * picking a head that matches their export.\n */\nexport type SegmenterHead = \"yolo-seg\";\n\nexport interface SegmenterOptions extends OrtSessionOptions {\n    /**\n     * Decoder family for the segmentation head. Default `\"yolo-seg\"` covers\n     * YOLOv8-seg/v11-seg/v26-seg.\n     */\n    readonly head?: SegmenterHead;\n    /** Class label spec — see {@link resolveLabels}. Defaults to the COCO 80-class preset. */\n    readonly labels?: LabelSpec;\n    /** Number of classes — used to validate the supplied labels. */\n    readonly numClasses?: number;\n    /** Model input `[width, height]` for letterboxing. Defaults to `[640, 640]`. */\n    readonly inputSize?: readonly [number, number];\n    /** Default minimum class score to keep a candidate. */\n    readonly confThreshold?: number;\n    /** Default IoU threshold for non-maximum suppression. */\n    readonly iouThreshold?: number;\n    /** Maximum number of instances per image. */\n    readonly maxDetections?: number;\n    /** Probability cutoff applied to soft masks. Defaults to `0.5`. */\n    readonly maskThreshold?: number;\n}\n\nexport interface SegmenterPredictOptions {\n    readonly confThreshold?: number;\n    readonly iouThreshold?: number;\n    /**\n     * If set, keep only instances whose `classId` is in this list.\n     * Mirrors Ultralytics' `model.predict(img, classes=[0, 16])`.\n     */\n    readonly classes?: readonly number[];\n}\n\n/**\n * Instance segmenter for YOLO seg ONNX models (v8-seg / v11-seg / ...).\n *\n * The model is expected to expose two outputs:\n *\n * 1. `output0`: `(1, 4 + numClasses + numMaskCoefs, numAnchors)` — per-anchor\n *    predictions (boxes, class scores, mask coefficients).\n * 2. `output1`: `(1, numMaskCoefs, maskH, maskW)` — prototype masks.\n *\n * `predict()` returns `Promise<SegmentationResults[]>` (length 1 for a\n * single image), mirroring Ultralytics' API. The envelope exposes:\n *\n * - `boxes`: bulk numpy view (`xyxy`, `xywh`, `xyxyn`, `xywhn`, `cls`, `conf`).\n * - `masks`: per-instance binary masks cropped to each box.\n * - per-instance {@link SegmentationResult} via iteration.\n *\n * @example\n * ```typescript\n * const seg = await Segmenter.create(\"/models/yolov8n-seg.onnx\");\n * const r = (await seg.predict(\"/images/street.jpg\"))[0];\n * for (const inst of r) {\n *   console.log(inst.cls, inst.conf, inst.box.xyxy);\n * }\n * ```\n */\nexport class Segmenter extends VisionTask {\n    private constructor(\n        session: OrtSession,\n        private readonly _head: SegmenterHead,\n        private readonly _labels: readonly string[],\n        private readonly _names: Readonly<Record<number, string>>,\n        private readonly _inputSize: readonly [number, number],\n        private readonly _confThreshold: number,\n        private readonly _iouThreshold: number,\n        private readonly _maxDetections: number,\n        private readonly _maskThreshold: number,\n    ) {\n        super(session);\n    }\n\n    /** Load the model and resolve labels. */\n    static async create(model: ModelSource, options: SegmenterOptions = {}): Promise<Segmenter> {\n        const head: SegmenterHead = options.head ?? \"yolo-seg\";\n        if (head !== \"yolo-seg\") {\n            throw new Error(`Unsupported segmenter head '${head}'. Supported: 'yolo-seg'.`);\n        }\n        const session = await OrtSession.create(model, options);\n        const labels = resolveLabels(options.labels ?? \"coco\", {\n            numClasses: options.numClasses,\n        });\n        const names: Record<number, string> = {};\n        for (let i = 0; i < labels.length; i++) {\n            names[i] = labels[i] as string;\n        }\n        return new Segmenter(\n            session,\n            head,\n            labels,\n            names,\n            options.inputSize ?? [640, 640],\n            options.confThreshold ?? 0.25,\n            options.iouThreshold ?? 0.45,\n            options.maxDetections ?? 300,\n            options.maskThreshold ?? 0.5,\n        );\n    }\n\n    /** The decoder family used to interpret the model's output. */\n    get head(): SegmenterHead {\n        return this._head;\n    }\n\n    /** Class labels indexed by class id. */\n    get labels(): readonly string[] {\n        return this._labels;\n    }\n\n    /** Class id → class name dict (matches Ultralytics' `model.names`). */\n    get names(): Readonly<Record<number, string>> {\n        return this._names;\n    }\n\n    /** Number of classes the model predicts. */\n    get numClasses(): number {\n        return this._labels.length;\n    }\n\n    /** Alias for {@link predict} (parity with PyTorch `nn.Module.__call__`). */\n    async call(\n        image: ImageInput,\n        options: SegmenterPredictOptions = {},\n    ): Promise<SegmentationResults[]> {\n        return this.predict(image, options);\n    }\n\n    /** Run instance segmentation on a single image. */\n    async predict(\n        image: ImageInput,\n        options: SegmenterPredictOptions = {},\n    ): Promise<SegmentationResults[]> {\n        const path = typeof image === \"string\" ? image : null;\n        const original = await loadImage(image);\n        const { tensor, scale, padLeft, padTop } = this._preprocess(original);\n        const outputs = await this._session.run({ [this._session.inputName]: tensor });\n\n        const { perAnchor, prototypes } = this._splitOutputs(outputs);\n\n        const decodedAll = decodeYoloSeg(\n            perAnchor.data as Float32Array,\n            perAnchor.dims,\n            prototypes.data as Float32Array,\n            prototypes.dims,\n            {\n                numClasses: this._labels.length,\n                inputWidth: this._inputSize[0],\n                inputHeight: this._inputSize[1],\n                originalWidth: original.width,\n                originalHeight: original.height,\n                padLeft,\n                padTop,\n                scale,\n                confThreshold: options.confThreshold ?? this._confThreshold,\n                iouThreshold: options.iouThreshold ?? this._iouThreshold,\n                maxDetections: this._maxDetections,\n                maskThreshold: this._maskThreshold,\n            },\n        );\n\n        const decoded =\n            options.classes !== undefined\n                ? (() => {\n                      const allowed = new Set(options.classes);\n                      return decodedAll.filter((d) => allowed.has(d.classId));\n                  })()\n                : decodedAll;\n\n        const detections = decoded.map((d) =>\n            this._buildResult(original, d.bbox, d.classId, d.confidence, d.mask),\n        );\n\n        const orig: readonly [number, number] = [original.height, original.width];\n        return [\n            new SegmentationResults(\n                this._buildBoxes(detections, orig),\n                this._buildMasks(detections, orig),\n                detections,\n                this._names,\n                original,\n                orig,\n                path,\n            ),\n        ];\n    }\n\n    private _preprocess(image: RGBImage): {\n        tensor: ort.Tensor;\n        scale: number;\n        padLeft: number;\n        padTop: number;\n    } {\n        const [tw, th] = this._inputSize;\n        const lb = letterbox(image, tw, th);\n        const f32 = toFloat32(lb.image);\n        const chw = toCHW(f32, lb.image.width, lb.image.height, 3);\n        return {\n            tensor: toFloat32Tensor(chw, [1, 3, lb.image.height, lb.image.width]),\n            scale: lb.scale,\n            padLeft: lb.padLeft,\n            padTop: lb.padTop,\n        };\n    }\n\n    private _splitOutputs(outputs: Record<string, ort.Tensor>): {\n        perAnchor: ort.Tensor;\n        prototypes: ort.Tensor;\n    } {\n        let perAnchor: ort.Tensor | undefined;\n        let prototypes: ort.Tensor | undefined;\n        for (const name of this._session.outputNames) {\n            const t = outputs[name];\n            if (t === undefined) continue;\n            if (t.dims.length === 3 && perAnchor === undefined) {\n                perAnchor = t;\n            } else if (t.dims.length === 4 && prototypes === undefined) {\n                prototypes = t;\n            }\n        }\n        if (perAnchor === undefined || prototypes === undefined) {\n            const shapes = this._session.outputNames.map(\n                (n) => `${n}: ${JSON.stringify(outputs[n]?.dims ?? [])}`,\n            );\n            throw new Error(\n                `Segmenter expected one 3-D and one 4-D output, got [${shapes.join(\", \")}].`,\n            );\n        }\n        return { perAnchor, prototypes };\n    }\n\n    private _buildResult(\n        original: RGBImage,\n        bbox: BoundingBox,\n        classId: number,\n        confidence: number,\n        mask: Mask,\n    ): SegmentationResult {\n        const [x1, y1, x2, y2] = bbox.asIntXyxy();\n        const cx1 = Math.max(0, x1);\n        const cy1 = Math.max(0, y1);\n        const cx2 = Math.min(original.width, x2);\n        const cy2 = Math.min(original.height, y2);\n\n        let segmentedImage: RGBImage;\n        let finalMask = mask;\n        if (cx2 > cx1 && cy2 > cy1 && mask.data.length > 0) {\n            const cropW = cx2 - cx1;\n            const cropH = cy2 - cy1;\n            const mw = Math.min(mask.width, cropW);\n            const mh = Math.min(mask.height, cropH);\n            const segData = new Uint8Array(mw * mh * 3);\n            for (let row = 0; row < mh; row++) {\n                const srcRowOffset = ((cy1 + row) * original.width + cx1) * 3;\n                const dstRowOffset = row * mw * 3;\n                const maskRowOffset = row * mask.width;\n                for (let col = 0; col < mw; col++) {\n                    const m = mask.data[maskRowOffset + col];\n                    if (m !== 0) {\n                        const s = srcRowOffset + col * 3;\n                        const d = dstRowOffset + col * 3;\n                        segData[d] = original.data[s];\n                        segData[d + 1] = original.data[s + 1];\n                        segData[d + 2] = original.data[s + 2];\n                    }\n                }\n            }\n            segmentedImage = new RGBImage(segData, mw, mh);\n            if (mw !== mask.width || mh !== mask.height) {\n                const trimmed = new Uint8Array(mw * mh);\n                for (let row = 0; row < mh; row++) {\n                    trimmed.set(\n                        mask.data.subarray(row * mask.width, row * mask.width + mw),\n                        row * mw,\n                    );\n                }\n                finalMask = new Mask(trimmed, mw, mh);\n            }\n        } else {\n            finalMask = new Mask(new Uint8Array(0), 0, 0);\n            segmentedImage = new RGBImage(new Uint8Array(0), 0, 0);\n        }\n\n        const className = this._names[classId] ?? `class_${classId}`;\n\n        return {\n            classId,\n            className,\n            confidence,\n            bbox,\n            cls: classId,\n            name: className,\n            conf: confidence,\n            box: bbox,\n            mask: finalMask,\n            segmentedImage,\n        };\n    }\n\n    private _buildBoxes(\n        detections: readonly SegmentationResult[],\n        origShape: readonly [number, number],\n    ): Boxes {\n        const n = detections.length;\n        const xyxy = new Float32Array(n * 4);\n        const cls = new Int32Array(n);\n        const conf = new Float32Array(n);\n        for (let i = 0; i < n; i++) {\n            const d = detections[i] as SegmentationResult;\n            xyxy[i * 4] = d.bbox.x1;\n            xyxy[i * 4 + 1] = d.bbox.y1;\n            xyxy[i * 4 + 2] = d.bbox.x2;\n            xyxy[i * 4 + 3] = d.bbox.y2;\n            cls[i] = d.classId;\n            conf[i] = d.confidence;\n        }\n        return new Boxes(xyxy, cls, conf, origShape);\n    }\n\n    private _buildMasks(\n        detections: readonly SegmentationResult[],\n        origShape: readonly [number, number],\n    ): Masks {\n        const xyxy = new Float32Array(detections.length * 4);\n        for (let i = 0; i < detections.length; i++) {\n            const d = detections[i] as SegmentationResult;\n            xyxy[i * 4] = d.bbox.x1;\n            xyxy[i * 4 + 1] = d.bbox.y1;\n            xyxy[i * 4 + 2] = d.bbox.x2;\n            xyxy[i * 4 + 3] = d.bbox.y2;\n        }\n        return new Masks(\n            detections.map((d) => d.mask),\n            xyxy,\n            origShape,\n        );\n    }\n}\n"],"mappings":";;;;;;;;;AAoFA,IAAa,IAAb,MAAa,UAAkB,EAAW;CAGjB;CACA;CACA;CACA;CACA;CACA;CACA;CACA;CATrB,YACI,GACA,GACA,GACA,GACA,GACA,GACA,GACA,GACA,GACF;EADmB,AAEjB,MAAM,CAAO,GATI,KAAA,QAAA,GACA,KAAA,UAAA,GACA,KAAA,SAAA,GACA,KAAA,aAAA,GACA,KAAA,iBAAA,GACA,KAAA,gBAAA,GACA,KAAA,iBAAA,GACA,KAAA,iBAAA;CAGrB;CAGA,aAAa,OAAO,GAAoB,IAA4B,CAAC,GAAuB;EACxF,IAAM,IAAsB,EAAQ,QAAQ;EAC5C,IAAI,MAAS,YACT,MAAU,MAAM,+BAA+B,EAAK,0BAA0B;EAElF,IAAM,IAAU,MAAM,EAAW,OAAO,GAAO,CAAO,GAChD,IAAS,EAAc,EAAQ,UAAU,QAAQ,EACnD,YAAY,EAAQ,WACxB,CAAC,GACK,IAAgC,CAAC;EACvC,KAAK,IAAI,IAAI,GAAG,IAAI,EAAO,QAAQ,KAC/B,EAAM,KAAK,EAAO;EAEtB,OAAO,IAAI,EACP,GACA,GACA,GACA,GACA,EAAQ,aAAa,CAAC,KAAK,GAAG,GAC9B,EAAQ,iBAAiB,KACzB,EAAQ,gBAAgB,KACxB,EAAQ,iBAAiB,KACzB,EAAQ,iBAAiB,EAC7B;CACJ;CAGA,IAAI,OAAsB;EACtB,OAAO,KAAK;CAChB;CAGA,IAAI,SAA4B;EAC5B,OAAO,KAAK;CAChB;CAGA,IAAI,QAA0C;EAC1C,OAAO,KAAK;CAChB;CAGA,IAAI,aAAqB;EACrB,OAAO,KAAK,QAAQ;CACxB;CAGA,MAAM,KACF,GACA,IAAmC,CAAC,GACN;EAC9B,OAAO,KAAK,QAAQ,GAAO,CAAO;CACtC;CAGA,MAAM,QACF,GACA,IAAmC,CAAC,GACN;EAC9B,IAAM,IAAO,OAAO,KAAU,WAAW,IAAQ,MAC3C,IAAW,MAAM,EAAU,CAAK,GAChC,EAAE,WAAQ,UAAO,YAAS,cAAW,KAAK,YAAY,CAAQ,GAC9D,IAAU,MAAM,KAAK,SAAS,IAAI,GAAG,KAAK,SAAS,YAAY,EAAO,CAAC,GAEvE,EAAE,cAAW,kBAAe,KAAK,cAAc,CAAO,GAEtD,IAAa,EACf,EAAU,MACV,EAAU,MACV,EAAW,MACX,EAAW,MACX;GACI,YAAY,KAAK,QAAQ;GACzB,YAAY,KAAK,WAAW;GAC5B,aAAa,KAAK,WAAW;GAC7B,eAAe,EAAS;GACxB,gBAAgB,EAAS;GACzB;GACA;GACA;GACA,eAAe,EAAQ,iBAAiB,KAAK;GAC7C,cAAc,EAAQ,gBAAgB,KAAK;GAC3C,eAAe,KAAK;GACpB,eAAe,KAAK;EACxB,CACJ,GAUM,KAPF,EAAQ,YAAY,KAAA,IAKd,WAJO;GACH,IAAM,IAAU,IAAI,IAAI,EAAQ,OAAO;GACvC,OAAO,EAAW,QAAQ,MAAM,EAAQ,IAAI,EAAE,OAAO,CAAC;EAC1D,EAAA,CAAG,EACH,CAEiB,KAAK,MAC5B,KAAK,aAAa,GAAU,EAAE,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,IAAI,CACvE,GAEM,IAAkC,CAAC,EAAS,QAAQ,EAAS,KAAK;EACxE,OAAO,CACH,IAAI,EACA,KAAK,YAAY,GAAY,CAAI,GACjC,KAAK,YAAY,GAAY,CAAI,GACjC,GACA,KAAK,QACL,GACA,GACA,CACJ,CACJ;CACJ;CAEA,YAAoB,GAKlB;EACE,IAAM,CAAC,GAAI,KAAM,KAAK,YAChB,IAAK,EAAU,GAAO,GAAI,CAAE;EAGlC,OAAO;GACH,QAAQ,EAFA,EADA,EAAU,EAAG,KACP,GAAK,EAAG,MAAM,OAAO,EAAG,MAAM,QAAQ,CAE5B,GAAK;IAAC;IAAG;IAAG,EAAG,MAAM;IAAQ,EAAG,MAAM;GAAK,CAAC;GACpE,OAAO,EAAG;GACV,SAAS,EAAG;GACZ,QAAQ,EAAG;EACf;CACJ;CAEA,cAAsB,GAGpB;EACE,IAAI,GACA;EACJ,KAAK,IAAM,KAAQ,KAAK,SAAS,aAAa;GAC1C,IAAM,IAAI,EAAQ;GACd,MAAM,KAAA,MACN,EAAE,KAAK,WAAW,KAAK,MAAc,KAAA,IACrC,IAAY,IACL,EAAE,KAAK,WAAW,KAAK,MAAe,KAAA,MAC7C,IAAa;EAErB;EACA,IAAI,MAAc,KAAA,KAAa,MAAe,KAAA,GAAW;GACrD,IAAM,IAAS,KAAK,SAAS,YAAY,KACpC,MAAM,GAAG,EAAE,IAAI,KAAK,UAAU,EAAQ,EAAE,EAAE,QAAQ,CAAC,CAAC,GACzD;GACA,MAAU,MACN,uDAAuD,EAAO,KAAK,IAAI,EAAE,GAC7E;EACJ;EACA,OAAO;GAAE;GAAW;EAAW;CACnC;CAEA,aACI,GACA,GACA,GACA,GACA,GACkB;EAClB,IAAM,CAAC,GAAI,GAAI,GAAI,KAAM,EAAK,UAAU,GAClC,IAAM,KAAK,IAAI,GAAG,CAAE,GACpB,IAAM,KAAK,IAAI,GAAG,CAAE,GACpB,IAAM,KAAK,IAAI,EAAS,OAAO,CAAE,GACjC,IAAM,KAAK,IAAI,EAAS,QAAQ,CAAE,GAEpC,GACA,IAAY;EAChB,IAAI,IAAM,KAAO,IAAM,KAAO,EAAK,KAAK,SAAS,GAAG;GAChD,IAAM,IAAQ,IAAM,GACd,IAAQ,IAAM,GACd,IAAK,KAAK,IAAI,EAAK,OAAO,CAAK,GAC/B,IAAK,KAAK,IAAI,EAAK,QAAQ,CAAK,GAChC,IAAU,IAAI,WAAW,IAAK,IAAK,CAAC;GAC1C,KAAK,IAAI,IAAM,GAAG,IAAM,GAAI,KAAO;IAC/B,IAAM,MAAiB,IAAM,KAAO,EAAS,QAAQ,KAAO,GACtD,IAAe,IAAM,IAAK,GAC1B,IAAgB,IAAM,EAAK;IACjC,KAAK,IAAI,IAAM,GAAG,IAAM,GAAI,KAExB,IADU,EAAK,KAAK,IAAgB,OAC1B,GAAG;KACT,IAAM,IAAI,IAAe,IAAM,GACzB,IAAI,IAAe,IAAM;KAG/B,AAFA,EAAQ,KAAK,EAAS,KAAK,IAC3B,EAAQ,IAAI,KAAK,EAAS,KAAK,IAAI,IACnC,EAAQ,IAAI,KAAK,EAAS,KAAK,IAAI;IACvC;GAER;GAEA,IADA,IAAiB,IAAI,EAAS,GAAS,GAAI,CAAE,GACzC,MAAO,EAAK,SAAS,MAAO,EAAK,QAAQ;IACzC,IAAM,IAAU,IAAI,WAAW,IAAK,CAAE;IACtC,KAAK,IAAI,IAAM,GAAG,IAAM,GAAI,KACxB,EAAQ,IACJ,EAAK,KAAK,SAAS,IAAM,EAAK,OAAO,IAAM,EAAK,QAAQ,CAAE,GAC1D,IAAM,CACV;IAEJ,IAAY,IAAI,EAAK,GAAS,GAAI,CAAE;GACxC;EACJ,OAEI,AADA,IAAY,IAAI,kBAAK,IAAI,WAAY,GAAG,GAAG,CAAC,GAC5C,IAAiB,IAAI,kBAAS,IAAI,WAAY,GAAG,GAAG,CAAC;EAGzD,IAAM,IAAY,KAAK,OAAO,MAAY,SAAS;EAEnD,OAAO;GACH;GACA;GACA;GACA;GACA,KAAK;GACL,MAAM;GACN,MAAM;GACN,KAAK;GACL,MAAM;GACN;EACJ;CACJ;CAEA,YACI,GACA,GACK;EACL,IAAM,IAAI,EAAW,QACf,IAAO,IAAI,aAAa,IAAI,CAAC,GAC7B,IAAM,IAAI,WAAW,CAAC,GACtB,IAAO,IAAI,aAAa,CAAC;EAC/B,KAAK,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK;GACxB,IAAM,IAAI,EAAW;GAMrB,AALA,EAAK,IAAI,KAAK,EAAE,KAAK,IACrB,EAAK,IAAI,IAAI,KAAK,EAAE,KAAK,IACzB,EAAK,IAAI,IAAI,KAAK,EAAE,KAAK,IACzB,EAAK,IAAI,IAAI,KAAK,EAAE,KAAK,IACzB,EAAI,KAAK,EAAE,SACX,EAAK,KAAK,EAAE;EAChB;EACA,OAAO,IAAI,EAAM,GAAM,GAAK,GAAM,CAAS;CAC/C;CAEA,YACI,GACA,GACK;EACL,IAAM,IAAO,IAAI,aAAa,EAAW,SAAS,CAAC;EACnD,KAAK,IAAI,IAAI,GAAG,IAAI,EAAW,QAAQ,KAAK;GACxC,IAAM,IAAI,EAAW;GAIrB,AAHA,EAAK,IAAI,KAAK,EAAE,KAAK,IACrB,EAAK,IAAI,IAAI,KAAK,EAAE,KAAK,IACzB,EAAK,IAAI,IAAI,KAAK,EAAE,KAAK,IACzB,EAAK,IAAI,IAAI,KAAK,EAAE,KAAK;EAC7B;EACA,OAAO,IAAI,EACP,EAAW,KAAK,MAAM,EAAE,IAAI,GAC5B,GACA,CACJ;CACJ;AACJ"}

package/dist/vision/types.cjs

@@ -1,2 +1,2 @@
-"use strict";Object.defineProperty(exports,Symbol.toStringTag,{value:"Module"});const r=require("./core/exceptions.cjs");class y{constructor(s,t,h){if(this.data=s,this.width=t,this.height=h,s.length!==t*h*3)throw new r.ImageLoadError(`RGBImage data length ${s.length} does not match width * height * 3 = ${t*h*3}.`)}data;width;height}class o{constructor(s,t,h,i){this.x1=s,this.y1=t,this.x2=h,this.y2=i}x1;y1;x2;y2;get width(){return Math.max(0,this.x2-this.x1)}get height(){return Math.max(0,this.y2-this.y1)}get area(){return this.width*this.height}get xyxy(){return[this.x1,this.y1,this.x2,this.y2]}get xywh(){return[(this.x1+this.x2)/2,(this.y1+this.y2)/2,this.width,this.height]}xyxyn(s){const[t,h]=s;return h<=0||t<=0?[0,0,0,0]:[this.x1/h,this.y1/t,this.x2/h,this.y2/t]}xywhn(s){const[t,h]=s;if(h<=0||t<=0)return[0,0,0,0];const[i,n,a,x]=this.xywh;return[i/h,n/t,a/h,x/t]}asXyxy(){return[this.x1,this.y1,this.x2,this.y2]}asXywh(){return[this.x1,this.y1,this.width,this.height]}asIntXyxy(){return[Math.trunc(this.x1),Math.trunc(this.y1),Math.trunc(this.x2),Math.trunc(this.y2)]}}class g{constructor(s,t,h){if(this.data=s,this.width=t,this.height=h,s.length!==t*h)throw new r.ImageLoadError(`Mask data length ${s.length} does not match width * height = ${t*h}.`)}data;width;height}exports.BoundingBox=o;exports.Mask=g;exports.RGBImage=y;
-//# sourceMappingURL=types.cjs.map
+const e=require("./core/exceptions.cjs");var t=class{data;width;height;constructor(t,n,r){if(this.data=t,this.width=n,this.height=r,t.length!==n*r*3)throw new e.ImageLoadError(`RGBImage data length ${t.length} does not match width * height * 3 = ${n*r*3}.`)}},n=class{x1;y1;x2;y2;constructor(e,t,n,r){this.x1=e,this.y1=t,this.x2=n,this.y2=r}get width(){return Math.max(0,this.x2-this.x1)}get height(){return Math.max(0,this.y2-this.y1)}get area(){return this.width*this.height}get xyxy(){return[this.x1,this.y1,this.x2,this.y2]}get xywh(){return[(this.x1+this.x2)/2,(this.y1+this.y2)/2,this.width,this.height]}xyxyn(e){let[t,n]=e;return n<=0||t<=0?[0,0,0,0]:[this.x1/n,this.y1/t,this.x2/n,this.y2/t]}xywhn(e){let[t,n]=e;if(n<=0||t<=0)return[0,0,0,0];let[r,i,a,o]=this.xywh;return[r/n,i/t,a/n,o/t]}asXyxy(){return[this.x1,this.y1,this.x2,this.y2]}asXywh(){return[this.x1,this.y1,this.width,this.height]}asIntXyxy(){return[Math.trunc(this.x1),Math.trunc(this.y1),Math.trunc(this.x2),Math.trunc(this.y2)]}},r=class{data;width;height;constructor(t,n,r){if(this.data=t,this.width=n,this.height=r,t.length!==n*r)throw new e.ImageLoadError(`Mask data length ${t.length} does not match width * height = ${n*r}.`)}};exports.BoundingBox=n,exports.Mask=r,exports.RGBImage=t;
+//# sourceMappingURL=types.cjs.map

package/dist/vision/types.cjs.map

@@ -1 +1 @@
-{"version":3,"file":"types.cjs","sources":["../../src/vision/types.ts"],"sourcesContent":["/**\n * Public output types returned by the SDK's vision tasks.\n *\n * These types form the contract between the SDK and its callers. They mirror\n * the Python `ort-vision-sdk` output dataclasses 1-to-1.\n *\n * Naming is intentionally compatible with the Ultralytics / torchvision idiom\n * (`cls`, `conf`, `box`, `xyxy`, `xywh`, normalized variants) so code ported\n * from those projects keeps working with minimal edits. The original verbose\n * names (`classId`, `className`, `confidence`, `bbox`) are still populated for\n * backwards compatibility.\n */\n\nimport { ImageLoadError } from \"./core/exceptions\";\n\n/**\n * HWC RGB uint8 image — the canonical image format used across the SDK.\n *\n * `data.length` must equal `width * height * 3`. The buffer is laid out row\n * by row, top-to-bottom, with each pixel as `[R, G, B]`.\n */\nexport class RGBImage {\n    constructor(\n        public readonly data: Uint8Array,\n        public readonly width: number,\n        public readonly height: number,\n    ) {\n        if (data.length !== width * height * 3) {\n            throw new ImageLoadError(\n                `RGBImage data length ${data.length} does not match width * height * 3 = ${\n                    width * height * 3\n                }.`,\n            );\n        }\n    }\n}\n\n/**\n * Axis-aligned bounding box in absolute pixel coordinates (xyxy format).\n *\n * Coordinates refer to the original input image (before any internal resize),\n * so callers can map detections back onto their source image without any\n * additional bookkeeping.\n */\nexport class BoundingBox {\n    constructor(\n        public readonly x1: number,\n        public readonly y1: number,\n        public readonly x2: number,\n        public readonly y2: number,\n    ) {}\n\n    /** Box width in pixels (clamped to non-negative). */\n    get width(): number {\n        return Math.max(0, this.x2 - this.x1);\n    }\n\n    /** Box height in pixels (clamped to non-negative). */\n    get height(): number {\n        return Math.max(0, this.y2 - this.y1);\n    }\n\n    /** Box area in pixels squared. */\n    get area(): number {\n        return this.width * this.height;\n    }\n\n    /** The box as `[x1, y1, x2, y2]` in absolute pixels (Ultralytics-style). */\n    get xyxy(): readonly [number, number, number, number] {\n        return [this.x1, this.y1, this.x2, this.y2];\n    }\n\n    /**\n     * The box as `[cx, cy, w, h]` with `(cx, cy)` at the center.\n     *\n     * Matches Ultralytics' `boxes.xywh` and YOLO's native head format. For the\n     * top-left `[x, y, w, h]` convention, use {@link asXywh}.\n     */\n    get xywh(): readonly [number, number, number, number] {\n        return [(this.x1 + this.x2) / 2, (this.y1 + this.y2) / 2, this.width, this.height];\n    }\n\n    /**\n     * The box as `[x1, y1, x2, y2]` normalized to `[0, 1]`.\n     *\n     * @param origShape `[height, width]` of the source image, in pixels.\n     */\n    xyxyn(origShape: readonly [number, number]): readonly [number, number, number, number] {\n        const [h, w] = origShape;\n        if (w <= 0 || h <= 0) return [0, 0, 0, 0];\n        return [this.x1 / w, this.y1 / h, this.x2 / w, this.y2 / h];\n    }\n\n    /**\n     * The box as `[cx, cy, w, h]` normalized to `[0, 1]`.\n     *\n     * @param origShape `[height, width]` of the source image, in pixels.\n     */\n    xywhn(origShape: readonly [number, number]): readonly [number, number, number, number] {\n        const [h, w] = origShape;\n        if (w <= 0 || h <= 0) return [0, 0, 0, 0];\n        const [cx, cy, bw, bh] = this.xywh;\n        return [cx / w, cy / h, bw / w, bh / h];\n    }\n\n    /** Returns `[x1, y1, x2, y2]`. */\n    asXyxy(): readonly [number, number, number, number] {\n        return [this.x1, this.y1, this.x2, this.y2];\n    }\n\n    /**\n     * Returns `[x, y, width, height]` with `(x, y)` at the **top-left**.\n     *\n     * Note: this is the top-left convention. Ultralytics' `xywh` getter uses\n     * **center** coordinates — for that, read {@link xywh}.\n     */\n    asXywh(): readonly [number, number, number, number] {\n        return [this.x1, this.y1, this.width, this.height];\n    }\n\n    /** Returns `[x1, y1, x2, y2]` truncated to integers, useful for slicing arrays. */\n    asIntXyxy(): readonly [number, number, number, number] {\n        return [Math.trunc(this.x1), Math.trunc(this.y1), Math.trunc(this.x2), Math.trunc(this.y2)];\n    }\n}\n\n/**\n * Probability assigned to a single class for a classification prediction.\n *\n * `cls` / `name` / `conf` are Ultralytics-style aliases populated alongside\n * the verbose `classId` / `className` / `probability` fields.\n */\nexport interface ClassProbability {\n    readonly classId: number;\n    readonly className: string;\n    readonly probability: number;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `probability` (Ultralytics-style). */\n    readonly conf: number;\n}\n\n/**\n * Output of an image classification inference.\n */\nexport interface ClassificationResult {\n    readonly classId: number;\n    readonly className: string;\n    readonly confidence: number;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `confidence` (Ultralytics-style). */\n    readonly conf: number;\n    /** The original input image as an HWC RGB uint8 array. */\n    readonly image: RGBImage;\n    /**\n     * Probabilities per class, sorted in descending order. The first entry\n     * mirrors `classId`, `className`, and `confidence`. When `topK` was passed\n     * to `predict`, the array is truncated to that length.\n     */\n    readonly probabilities: readonly ClassProbability[];\n}\n\n/**\n * Single detected object produced by an object-detection model.\n */\nexport interface DetectionResult {\n    readonly classId: number;\n    readonly className: string;\n    readonly confidence: number;\n    readonly bbox: BoundingBox;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `confidence` (Ultralytics-style). */\n    readonly conf: number;\n    /** Alias for `bbox` (Ultralytics-style). */\n    readonly box: BoundingBox;\n    /**\n     * The original image cropped to `bbox`, HWC RGB uint8. Empty boxes\n     * (zero area) yield a zero-sized `RGBImage`.\n     */\n    readonly croppedImage: RGBImage;\n}\n\n/**\n * Single-channel binary or grayscale mask, laid out row-major.\n *\n * `data.length` must equal `width * height`. For binary masks, values are\n * `0` (background) or `255` (foreground); soft masks may use the full\n * `[0, 255]` range.\n */\nexport class Mask {\n    constructor(\n        public readonly data: Uint8Array,\n        public readonly width: number,\n        public readonly height: number,\n    ) {\n        if (data.length !== width * height) {\n            throw new ImageLoadError(\n                `Mask data length ${data.length} does not match width * height = ${\n                    width * height\n                }.`,\n            );\n        }\n    }\n}\n\n/**\n * Single segmented instance produced by an instance-segmentation model.\n *\n * Mirrors {@link DetectionResult} and adds the per-instance binary mask\n * plus a \"ready-to-display\" background-removed crop.\n */\nexport interface SegmentationResult {\n    readonly classId: number;\n    readonly className: string;\n    readonly confidence: number;\n    readonly bbox: BoundingBox;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `confidence` (Ultralytics-style). */\n    readonly conf: number;\n    /** Alias for `bbox` (Ultralytics-style). */\n    readonly box: BoundingBox;\n    /**\n     * Binary mask cropped to `bbox`. Values are `0` (background) or `255`\n     * (foreground). Empty boxes yield a zero-sized `Mask`.\n     */\n    readonly mask: Mask;\n    /**\n     * The original image cropped to `bbox` with background pixels (where\n     * `mask.data[i] === 0`) zeroed out. Empty boxes yield a zero-sized\n     * `RGBImage`.\n     */\n    readonly segmentedImage: RGBImage;\n}\n"],"names":["RGBImage","data","width","height","ImageLoadError","BoundingBox","x1","y1","x2","y2","origShape","h","w","cx","cy","bw","bh","Mask"],"mappings":"yHAqBO,MAAMA,CAAS,CAClB,YACoBC,EACAC,EACAC,EAClB,CACE,GAJgB,KAAA,KAAAF,EACA,KAAA,MAAAC,EACA,KAAA,OAAAC,EAEZF,EAAK,SAAWC,EAAQC,EAAS,EACjC,MAAM,IAAIC,EAAAA,eACN,wBAAwBH,EAAK,MAAM,wCAC/BC,EAAQC,EAAS,CACrB,GAAA,CAGZ,CAXoB,KACA,MACA,MAUxB,CASO,MAAME,CAAY,CACrB,YACoBC,EACAC,EACAC,EACAC,EAClB,CAJkB,KAAA,GAAAH,EACA,KAAA,GAAAC,EACA,KAAA,GAAAC,EACA,KAAA,GAAAC,CACjB,CAJiB,GACA,GACA,GACA,GAIpB,IAAI,OAAgB,CAChB,OAAO,KAAK,IAAI,EAAG,KAAK,GAAK,KAAK,EAAE,CACxC,CAGA,IAAI,QAAiB,CACjB,OAAO,KAAK,IAAI,EAAG,KAAK,GAAK,KAAK,EAAE,CACxC,CAGA,IAAI,MAAe,CACf,OAAO,KAAK,MAAQ,KAAK,MAC7B,CAGA,IAAI,MAAkD,CAClD,MAAO,CAAC,KAAK,GAAI,KAAK,GAAI,KAAK,GAAI,KAAK,EAAE,CAC9C,CAQA,IAAI,MAAkD,CAClD,MAAO,EAAE,KAAK,GAAK,KAAK,IAAM,GAAI,KAAK,GAAK,KAAK,IAAM,EAAG,KAAK,MAAO,KAAK,MAAM,CACrF,CAOA,MAAMC,EAAiF,CACnF,KAAM,CAACC,EAAGC,CAAC,EAAIF,EACf,OAAIE,GAAK,GAAKD,GAAK,EAAU,CAAC,EAAG,EAAG,EAAG,CAAC,EACjC,CAAC,KAAK,GAAKC,EAAG,KAAK,GAAKD,EAAG,KAAK,GAAKC,EAAG,KAAK,GAAKD,CAAC,CAC9D,CAOA,MAAMD,EAAiF,CACnF,KAAM,CAACC,EAAGC,CAAC,EAAIF,EACf,GAAIE,GAAK,GAAKD,GAAK,QAAU,CAAC,EAAG,EAAG,EAAG,CAAC,EACxC,KAAM,CAACE,EAAIC,EAAIC,EAAIC,CAAE,EAAI,KAAK,KAC9B,MAAO,CAACH,EAAKD,EAAGE,EAAKH,EAAGI,EAAKH,EAAGI,EAAKL,CAAC,CAC1C,CAGA,QAAoD,CAChD,MAAO,CAAC,KAAK,GAAI,KAAK,GAAI,KAAK,GAAI,KAAK,EAAE,CAC9C,CAQA,QAAoD,CAChD,MAAO,CAAC,KAAK,GAAI,KAAK,GAAI,KAAK,MAAO,KAAK,MAAM,CACrD,CAGA,WAAuD,CACnD,MAAO,CAAC,KAAK,MAAM,KAAK,EAAE,EAAG,KAAK,MAAM,KAAK,EAAE,EAAG,KAAK,MAAM,KAAK,EAAE,EAAG,KAAK,MAAM,KAAK,EAAE,CAAC,CAC9F,CACJ,CAyEO,MAAMM,CAAK,CACd,YACoBhB,EACAC,EACAC,EAClB,CACE,GAJgB,KAAA,KAAAF,EACA,KAAA,MAAAC,EACA,KAAA,OAAAC,EAEZF,EAAK,SAAWC,EAAQC,EACxB,MAAM,IAAIC,EAAAA,eACN,oBAAoBH,EAAK,MAAM,oCAC3BC,EAAQC,CACZ,GAAA,CAGZ,CAXoB,KACA,MACA,MAUxB"}
+{"version":3,"file":"types.cjs","names":[],"sources":["../../src/vision/types.ts"],"sourcesContent":["/**\n * Public output types returned by the SDK's vision tasks.\n *\n * These types form the contract between the SDK and its callers. They mirror\n * the Python `ort-vision-sdk` output dataclasses 1-to-1.\n *\n * Naming is intentionally compatible with the Ultralytics / torchvision idiom\n * (`cls`, `conf`, `box`, `xyxy`, `xywh`, normalized variants) so code ported\n * from those projects keeps working with minimal edits. The original verbose\n * names (`classId`, `className`, `confidence`, `bbox`) are still populated for\n * backwards compatibility.\n */\n\nimport { ImageLoadError } from \"./core/exceptions\";\n\n/**\n * HWC RGB uint8 image — the canonical image format used across the SDK.\n *\n * `data.length` must equal `width * height * 3`. The buffer is laid out row\n * by row, top-to-bottom, with each pixel as `[R, G, B]`.\n */\nexport class RGBImage {\n    constructor(\n        public readonly data: Uint8Array,\n        public readonly width: number,\n        public readonly height: number,\n    ) {\n        if (data.length !== width * height * 3) {\n            throw new ImageLoadError(\n                `RGBImage data length ${data.length} does not match width * height * 3 = ${\n                    width * height * 3\n                }.`,\n            );\n        }\n    }\n}\n\n/**\n * Axis-aligned bounding box in absolute pixel coordinates (xyxy format).\n *\n * Coordinates refer to the original input image (before any internal resize),\n * so callers can map detections back onto their source image without any\n * additional bookkeeping.\n */\nexport class BoundingBox {\n    constructor(\n        public readonly x1: number,\n        public readonly y1: number,\n        public readonly x2: number,\n        public readonly y2: number,\n    ) {}\n\n    /** Box width in pixels (clamped to non-negative). */\n    get width(): number {\n        return Math.max(0, this.x2 - this.x1);\n    }\n\n    /** Box height in pixels (clamped to non-negative). */\n    get height(): number {\n        return Math.max(0, this.y2 - this.y1);\n    }\n\n    /** Box area in pixels squared. */\n    get area(): number {\n        return this.width * this.height;\n    }\n\n    /** The box as `[x1, y1, x2, y2]` in absolute pixels (Ultralytics-style). */\n    get xyxy(): readonly [number, number, number, number] {\n        return [this.x1, this.y1, this.x2, this.y2];\n    }\n\n    /**\n     * The box as `[cx, cy, w, h]` with `(cx, cy)` at the center.\n     *\n     * Matches Ultralytics' `boxes.xywh` and YOLO's native head format. For the\n     * top-left `[x, y, w, h]` convention, use {@link asXywh}.\n     */\n    get xywh(): readonly [number, number, number, number] {\n        return [(this.x1 + this.x2) / 2, (this.y1 + this.y2) / 2, this.width, this.height];\n    }\n\n    /**\n     * The box as `[x1, y1, x2, y2]` normalized to `[0, 1]`.\n     *\n     * @param origShape `[height, width]` of the source image, in pixels.\n     */\n    xyxyn(origShape: readonly [number, number]): readonly [number, number, number, number] {\n        const [h, w] = origShape;\n        if (w <= 0 || h <= 0) return [0, 0, 0, 0];\n        return [this.x1 / w, this.y1 / h, this.x2 / w, this.y2 / h];\n    }\n\n    /**\n     * The box as `[cx, cy, w, h]` normalized to `[0, 1]`.\n     *\n     * @param origShape `[height, width]` of the source image, in pixels.\n     */\n    xywhn(origShape: readonly [number, number]): readonly [number, number, number, number] {\n        const [h, w] = origShape;\n        if (w <= 0 || h <= 0) return [0, 0, 0, 0];\n        const [cx, cy, bw, bh] = this.xywh;\n        return [cx / w, cy / h, bw / w, bh / h];\n    }\n\n    /** Returns `[x1, y1, x2, y2]`. */\n    asXyxy(): readonly [number, number, number, number] {\n        return [this.x1, this.y1, this.x2, this.y2];\n    }\n\n    /**\n     * Returns `[x, y, width, height]` with `(x, y)` at the **top-left**.\n     *\n     * Note: this is the top-left convention. Ultralytics' `xywh` getter uses\n     * **center** coordinates — for that, read {@link xywh}.\n     */\n    asXywh(): readonly [number, number, number, number] {\n        return [this.x1, this.y1, this.width, this.height];\n    }\n\n    /** Returns `[x1, y1, x2, y2]` truncated to integers, useful for slicing arrays. */\n    asIntXyxy(): readonly [number, number, number, number] {\n        return [Math.trunc(this.x1), Math.trunc(this.y1), Math.trunc(this.x2), Math.trunc(this.y2)];\n    }\n}\n\n/**\n * Probability assigned to a single class for a classification prediction.\n *\n * `cls` / `name` / `conf` are Ultralytics-style aliases populated alongside\n * the verbose `classId` / `className` / `probability` fields.\n */\nexport interface ClassProbability {\n    readonly classId: number;\n    readonly className: string;\n    readonly probability: number;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `probability` (Ultralytics-style). */\n    readonly conf: number;\n}\n\n/**\n * Output of an image classification inference.\n */\nexport interface ClassificationResult {\n    readonly classId: number;\n    readonly className: string;\n    readonly confidence: number;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `confidence` (Ultralytics-style). */\n    readonly conf: number;\n    /** The original input image as an HWC RGB uint8 array. */\n    readonly image: RGBImage;\n    /**\n     * Probabilities per class, sorted in descending order. The first entry\n     * mirrors `classId`, `className`, and `confidence`. When `topK` was passed\n     * to `predict`, the array is truncated to that length.\n     */\n    readonly probabilities: readonly ClassProbability[];\n}\n\n/**\n * Single detected object produced by an object-detection model.\n */\nexport interface DetectionResult {\n    readonly classId: number;\n    readonly className: string;\n    readonly confidence: number;\n    readonly bbox: BoundingBox;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `confidence` (Ultralytics-style). */\n    readonly conf: number;\n    /** Alias for `bbox` (Ultralytics-style). */\n    readonly box: BoundingBox;\n    /**\n     * The original image cropped to `bbox`, HWC RGB uint8. Empty boxes\n     * (zero area) yield a zero-sized `RGBImage`.\n     */\n    readonly croppedImage: RGBImage;\n}\n\n/**\n * Single-channel binary or grayscale mask, laid out row-major.\n *\n * `data.length` must equal `width * height`. For binary masks, values are\n * `0` (background) or `255` (foreground); soft masks may use the full\n * `[0, 255]` range.\n */\nexport class Mask {\n    constructor(\n        public readonly data: Uint8Array,\n        public readonly width: number,\n        public readonly height: number,\n    ) {\n        if (data.length !== width * height) {\n            throw new ImageLoadError(\n                `Mask data length ${data.length} does not match width * height = ${\n                    width * height\n                }.`,\n            );\n        }\n    }\n}\n\n/**\n * Single segmented instance produced by an instance-segmentation model.\n *\n * Mirrors {@link DetectionResult} and adds the per-instance binary mask\n * plus a \"ready-to-display\" background-removed crop.\n */\nexport interface SegmentationResult {\n    readonly classId: number;\n    readonly className: string;\n    readonly confidence: number;\n    readonly bbox: BoundingBox;\n    /** Alias for `classId` (Ultralytics-style). */\n    readonly cls: number;\n    /** Alias for `className`. */\n    readonly name: string;\n    /** Alias for `confidence` (Ultralytics-style). */\n    readonly conf: number;\n    /** Alias for `bbox` (Ultralytics-style). */\n    readonly box: BoundingBox;\n    /**\n     * Binary mask cropped to `bbox`. Values are `0` (background) or `255`\n     * (foreground). Empty boxes yield a zero-sized `Mask`.\n     */\n    readonly mask: Mask;\n    /**\n     * The original image cropped to `bbox` with background pixels (where\n     * `mask.data[i] === 0`) zeroed out. Empty boxes yield a zero-sized\n     * `RGBImage`.\n     */\n    readonly segmentedImage: RGBImage;\n}\n"],"mappings":"yCAqBA,IAAa,EAAb,KAAsB,CAEE,KACA,MACA,OAHpB,YACI,EACA,EACA,EACF,CACE,GAJgB,KAAA,KAAA,EACA,KAAA,MAAA,EACA,KAAA,OAAA,EAEZ,EAAK,SAAW,EAAQ,EAAS,EACjC,MAAM,IAAI,EAAA,eACN,wBAAwB,EAAK,OAAO,uCAChC,EAAQ,EAAS,EACpB,EACL,CAER,CACJ,EASa,EAAb,KAAyB,CAED,GACA,GACA,GACA,GAJpB,YACI,EACA,EACA,EACA,EACF,CAJkB,KAAA,GAAA,EACA,KAAA,GAAA,EACA,KAAA,GAAA,EACA,KAAA,GAAA,CACjB,CAGH,IAAI,OAAgB,CAChB,OAAO,KAAK,IAAI,EAAG,KAAK,GAAK,KAAK,EAAE,CACxC,CAGA,IAAI,QAAiB,CACjB,OAAO,KAAK,IAAI,EAAG,KAAK,GAAK,KAAK,EAAE,CACxC,CAGA,IAAI,MAAe,CACf,OAAO,KAAK,MAAQ,KAAK,MAC7B,CAGA,IAAI,MAAkD,CAClD,MAAO,CAAC,KAAK,GAAI,KAAK,GAAI,KAAK,GAAI,KAAK,EAAE,CAC9C,CAQA,IAAI,MAAkD,CAClD,MAAO,EAAE,KAAK,GAAK,KAAK,IAAM,GAAI,KAAK,GAAK,KAAK,IAAM,EAAG,KAAK,MAAO,KAAK,MAAM,CACrF,CAOA,MAAM,EAAiF,CACnF,GAAM,CAAC,EAAG,GAAK,EAEf,OADI,GAAK,GAAK,GAAK,EAAU,CAAC,EAAG,EAAG,EAAG,CAAC,EACjC,CAAC,KAAK,GAAK,EAAG,KAAK,GAAK,EAAG,KAAK,GAAK,EAAG,KAAK,GAAK,CAAC,CAC9D,CAOA,MAAM,EAAiF,CACnF,GAAM,CAAC,EAAG,GAAK,EACf,GAAI,GAAK,GAAK,GAAK,EAAG,MAAO,CAAC,EAAG,EAAG,EAAG,CAAC,EACxC,GAAM,CAAC,EAAI,EAAI,EAAI,GAAM,KAAK,KAC9B,MAAO,CAAC,EAAK,EAAG,EAAK,EAAG,EAAK,EAAG,EAAK,CAAC,CAC1C,CAGA,QAAoD,CAChD,MAAO,CAAC,KAAK,GAAI,KAAK,GAAI,KAAK,GAAI,KAAK,EAAE,CAC9C,CAQA,QAAoD,CAChD,MAAO,CAAC,KAAK,GAAI,KAAK,GAAI,KAAK,MAAO,KAAK,MAAM,CACrD,CAGA,WAAuD,CACnD,MAAO,CAAC,KAAK,MAAM,KAAK,EAAE,EAAG,KAAK,MAAM,KAAK,EAAE,EAAG,KAAK,MAAM,KAAK,EAAE,EAAG,KAAK,MAAM,KAAK,EAAE,CAAC,CAC9F,CACJ,EAyEa,EAAb,KAAkB,CAEM,KACA,MACA,OAHpB,YACI,EACA,EACA,EACF,CACE,GAJgB,KAAA,KAAA,EACA,KAAA,MAAA,EACA,KAAA,OAAA,EAEZ,EAAK,SAAW,EAAQ,EACxB,MAAM,IAAI,EAAA,eACN,oBAAoB,EAAK,OAAO,mCAC5B,EAAQ,EACX,EACL,CAER,CACJ"}