first commit

2026-10-06 12:16:46 +09:00
commit 619d980098
415 changed files with 22283 additions and 0 deletions
@@ -0,0 +1,141 @@
+import { pipeline, type DepthEstimationPipelineOutput } from '@huggingface/transformers';
+
+const MODEL_ID = 'onnx-community/depth-anything-v2-small-ONNX';
+type Device = 'webgpu' | 'wasm';
+type RequestMessage = { type: 'infer'; id: number; image: ImageBitmap };
+type WorkerReply =
+  | { type: 'status'; id: number; message: string }
+  | { type: 'result'; id: number; width: number; height: number; values: ArrayBuffer }
+  | { type: 'error'; id: number; message: string };
+type Estimator = {
+  (image: OffscreenCanvas): Promise<DepthEstimationPipelineOutput | DepthEstimationPipelineOutput[]>;
+  dispose(): Promise<void>;
+};
+type Progress = { status: string; progress?: number };
+type PipelineFactory = (task: string, model: string, options: {
+  device: Device;
+  progress_callback: (progress: Progress) => void;
+}) => Promise<Estimator>;
+const createPipeline = pipeline as unknown as PipelineFactory;
+const workerScope = self as unknown as {
+  postMessage(message: WorkerReply, transfer?: Transferable[]): void;
+  addEventListener(type: 'message', listener: (event: MessageEvent<RequestMessage>) => void): void;
+};
+
+let estimator: Estimator | null = null;
+let activeDevice: Device | null = null;
+
+function reply(message: WorkerReply, transfer: Transferable[] = []): void {
+  workerScope.postMessage(message, transfer);
+}
+
+function setStatus(id: number, message: string): void {
+  reply({ type: 'status', id, message });
+}
+
+async function loadEstimator(device: Device, id: number): Promise<Estimator> {
+  setStatus(id, 'AI 모델 준비 중…');
+  return createPipeline('depth-estimation', MODEL_ID, {
+    device,
+    progress_callback: (progress) => {
+      if (progress.status === 'progress') {
+        setStatus(id, `AI 모델 다운로드 중 · ${Math.round(progress.progress ?? 0)}%`);
+      } else if (progress.status === 'initiate' || progress.status === 'download') {
+        setStatus(id, 'AI 모델 다운로드 중…');
+      }
+    },
+  });
+}
+
+async function ensureEstimator(id: number): Promise<Estimator> {
+  if (estimator) return estimator;
+  if ('gpu' in navigator) {
+    try {
+      estimator = await loadEstimator('webgpu', id);
+      activeDevice = 'webgpu';
+      return estimator;
+    } catch {
+      if (estimator) await estimator.dispose();
+      estimator = null;
+      setStatus(id, 'AI 모델을 준비하는 중입니다…');
+    }
+  }
+  estimator = await loadEstimator('wasm', id);
+  activeDevice = 'wasm';
+  return estimator;
+}
+
+function normalizeDepth(output: DepthEstimationPipelineOutput): { values: Uint8Array; width: number; height: number } {
+  const { predicted_depth: prediction } = output;
+  const [height, width] = prediction.dims.slice(-2);
+  const data = prediction.data;
+  if (!width || !height || data.length !== width * height) throw new Error('모델이 유효한 depth 크기를 반환하지 않았습니다.');
+
+  let min = Infinity;
+  let max = -Infinity;
+  for (let i = 0; i < data.length; i += 1) {
+    const value = Number(data[i]);
+    if (!Number.isFinite(value)) continue;
+    min = Math.min(min, value);
+    max = Math.max(max, value);
+  }
+  if (!Number.isFinite(min) || !Number.isFinite(max)) throw new Error('모델의 depth 결과가 비어 있습니다.');
+
+  const range = max - min;
+  const values = new Uint8Array(width * height);
+  if (range <= 1e-8) {
+    values.fill(128);
+  } else {
+    for (let i = 0; i < values.length; i += 1) {
+      const value = Number(data[i]);
+      // Depth Anything V2 predicts relative inverse depth: larger values are nearer.
+      values[i] = Number.isFinite(value) ? Math.round(((value - min) / range) * 255) : 128;
+    }
+  }
+  return { values, width, height };
+}
+
+async function infer({ id, image }: RequestMessage): Promise<void> {
+  try {
+    const canvas = new OffscreenCanvas(image.width, image.height);
+    const context = canvas.getContext('2d');
+    if (!context) throw new Error('사진을 처리할 수 없습니다.');
+    context.drawImage(image, 0, 0);
+    image.close();
+
+    let model = await ensureEstimator(id);
+    setStatus(id, 'AI 깊이를 계산하는 중입니다…');
+    let output: DepthEstimationPipelineOutput;
+    try {
+      const predicted = await model(canvas);
+      if (Array.isArray(predicted)) throw new Error('단일 사진 결과를 읽을 수 없습니다.');
+      output = predicted;
+    } catch (error) {
+      if (activeDevice !== 'webgpu') throw error;
+      await model.dispose();
+      estimator = null;
+      activeDevice = null;
+      setStatus(id, 'AI 추론을 다시 시도하는 중입니다…');
+      model = await loadEstimator('wasm', id);
+      estimator = model;
+      activeDevice = 'wasm';
+      const predicted = await model(canvas);
+      if (Array.isArray(predicted)) throw new Error('단일 사진 결과를 읽을 수 없습니다.');
+      output = predicted;
+    }
+
+    try {
+      const normalized = normalizeDepth(output);
+      const values = normalized.values.buffer as ArrayBuffer;
+      reply({ type: 'result', id, width: normalized.width, height: normalized.height, values }, [values]);
+    } finally {
+      output.predicted_depth.dispose();
+    }
+  } catch (error) {
+    reply({ type: 'error', id, message: error instanceof Error ? error.message : 'AI depth 생성에 실패했습니다.' });
+  }
+}
+
+workerScope.addEventListener('message', (event: MessageEvent<RequestMessage>) => {
+  if (event.data.type === 'infer') void infer(event.data);
+});