Initial commit: digital reception monorepo (M1-M11 + demo extensions)

2026-05-25 12:59:54 +05:00
commit b9f88194d9
182 changed files with 20578 additions and 0 deletions
@@ -0,0 +1,50 @@
+"""CLI: python -m video_ingest --source clip.mp4 --camera-name cam-A --zone A."""
+from __future__ import annotations
+
+import argparse
+import logging
+from pathlib import Path
+
+from dotenv import load_dotenv
+
+from .ingestor import IngestConfig, run
+
+load_dotenv(Path(__file__).parent.parent.parent.parent / ".env")
+load_dotenv()
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(prog="video_ingest")
+    parser.add_argument("--source", required=True, type=Path, help="Путь к mp4 файлу")
+    parser.add_argument(
+        "--camera-name",
+        required=True,
+        help="Имя камеры (должно совпадать с seeded — cam-entrance/cam-corridor/cam-reception)",
+    )
+    parser.add_argument(
+        "--zone",
+        choices=["A", "B", "C"],
+        help="Код зоны (опционально — будет взят из связки камеры)",
+    )
+    parser.add_argument("--sample-every", type=int, default=None)
+    parser.add_argument("--log-level", default="INFO")
+    args = parser.parse_args()
+
+    logging.basicConfig(
+        level=args.log_level,
+        format="%(asctime)s %(levelname)s [%(name)s] %(message)s",
+    )
+
+    cfg = IngestConfig(
+        source=args.source,
+        camera_name=args.camera_name,
+        zone_code=args.zone,
+        **({"sample_every": args.sample_every} if args.sample_every else {}),
+    )
+
+    result = run(cfg)
+    print(result)
+
+
+if __name__ == "__main__":
+    main()
@@ -0,0 +1,82 @@
+"""HTTP клиенты к face-service и apps/api для минимального video-ingest."""
+from __future__ import annotations
+
+import os
+import base64
+import logging
+from datetime import datetime
+from typing import Any
+
+import requests
+
+logger = logging.getLogger(__name__)
+
+
+class FaceServiceClient:
+    def __init__(self, base_url: str | None = None) -> None:
+        self.base_url = (base_url or os.getenv("FACE_SERVICE_URL", "http://localhost:8001")).rstrip("/")
+
+    def embed(self, jpeg_bytes: bytes) -> tuple[list[float], float] | None:
+        """Возвращает (embedding, quality) или None если лицо не найдено."""
+        b64 = base64.b64encode(jpeg_bytes).decode("ascii")
+        r = requests.post(f"{self.base_url}/embed", json={"frame": b64}, timeout=10)
+        r.raise_for_status()
+        data = r.json()
+        if data is None:
+            return None
+        return data["embedding"], data["quality"]
+
+    def save_track_embedding(
+        self,
+        jpeg_bytes: bytes,
+        track_id: str,
+        camera_id: str,
+        captured_at: datetime,
+    ) -> dict[str, Any] | None:
+        b64 = base64.b64encode(jpeg_bytes).decode("ascii")
+        r = requests.post(
+            f"{self.base_url}/track-embeddings",
+            json={
+                "frame": b64,
+                "track_id": track_id,
+                "camera_id": camera_id,
+                "captured_at": captured_at.isoformat(),
+            },
+            timeout=10,
+        )
+        r.raise_for_status()
+        return r.json()
+
+
+class ReceptionApiClient:
+    def __init__(self, base_url: str | None = None) -> None:
+        self.base_url = (base_url or os.getenv("API_BASE_URL", "http://localhost:4000")).rstrip("/")
+
+    def create_track(self, camera_name: str, first_seen_at: datetime) -> dict[str, Any]:
+        r = requests.post(
+            f"{self.base_url}/ingest/tracks",
+            json={"cameraName": camera_name, "firstSeenAt": first_seen_at.isoformat()},
+            timeout=10,
+        )
+        r.raise_for_status()
+        return r.json()
+
+    def add_event(
+        self,
+        track_id: str,
+        event_type: str,
+        camera_name: str,
+        occurred_at: datetime,
+        zone_code: str | None = None,
+    ) -> dict[str, Any]:
+        body = {
+            "trackId": track_id,
+            "type": event_type,
+            "cameraName": camera_name,
+            "occurredAt": occurred_at.isoformat(),
+        }
+        if zone_code:
+            body["zoneCode"] = zone_code
+        r = requests.post(f"{self.base_url}/ingest/track-events", json=body, timeout=10)
+        r.raise_for_status()
+        return r.json()
@@ -0,0 +1,144 @@
+"""Минимальный конвейер mp4 → face-service /embed → tracks → apps/api.
+
+Замысел: smoke-тест для цепочки «видео → эмбеддинг → трек → БД».
+Это НЕ боевой video-ingest из ТЗ §6 (RTSP, ByteTrack, GPU) — это
+скаффолд Фазы 1, чтобы Фаза 0 не блокировала остальную работу.
+
+Алгоритм:
+  1. Открыть mp4 через OpenCV.
+  2. Для каждого N-го кадра (SAMPLE_EVERY): JPEG → face-service /embed.
+  3. Если эмбеддинг есть — single-camera трекинг по cos-дистанции:
+     если дистанция к last_embedding < TRACK_DISTANCE_THRESH и
+     прошло < TRACK_WINDOW_SEC секунд → тот же трек.
+     Иначе — новый трек (создаём через apps/api /ingest/tracks).
+  4. Сохраняем эмбеддинг через face-service /track-embeddings.
+  5. На первом кадре трека постим событие `arrived` через /ingest/track-events.
+"""
+from __future__ import annotations
+
+import logging
+import os
+from dataclasses import dataclass
+from datetime import datetime, timedelta
+from pathlib import Path
+
+import cv2
+import numpy as np
+
+from .api import FaceServiceClient, ReceptionApiClient
+
+logger = logging.getLogger(__name__)
+
+
+@dataclass
+class IngestConfig:
+    source: Path
+    camera_name: str
+    zone_code: str | None = None
+    sample_every: int = int(os.getenv("SAMPLE_EVERY", "10"))
+    track_distance_thresh: float = float(os.getenv("TRACK_DISTANCE_THRESH", "0.3"))
+    track_window_sec: float = float(os.getenv("TRACK_WINDOW_SEC", "2.0"))
+    jpeg_quality: int = 85
+    real_time_start: datetime | None = None
+
+
+def cosine_distance(a: np.ndarray, b: np.ndarray) -> float:
+    return float(1.0 - np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b)))
+
+
+def run(cfg: IngestConfig) -> dict:
+    cap = cv2.VideoCapture(str(cfg.source))
+    if not cap.isOpened():
+        raise RuntimeError(f"Не могу открыть видео: {cfg.source}")
+
+    fps = cap.get(cv2.CAP_PROP_FPS) or 25.0
+    total = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    logger.info(f"Открыто видео {cfg.source} ({total} кадров, {fps:.1f} fps)")
+
+    face = FaceServiceClient()
+    api = ReceptionApiClient()
+    base_time = cfg.real_time_start or datetime.utcnow()
+
+    current_track_id: str | None = None
+    current_track_camera_id: str | None = None
+    last_embedding: np.ndarray | None = None
+    last_capture_time: datetime | None = None
+    tracks_created = 0
+    embeddings_saved = 0
+    frame_idx = 0
+
+    while True:
+        ok, frame = cap.read()
+        if not ok:
+            break
+        frame_idx += 1
+        if frame_idx % cfg.sample_every != 0:
+            continue
+
+        captured_at = base_time + timedelta(seconds=frame_idx / fps)
+
+        ok_jpg, buf = cv2.imencode(".jpg", frame, [int(cv2.IMWRITE_JPEG_QUALITY), cfg.jpeg_quality])
+        if not ok_jpg:
+            logger.warning(f"frame {frame_idx}: jpeg encode failed")
+            continue
+        jpeg_bytes = buf.tobytes()
+
+        try:
+            embed_result = face.embed(jpeg_bytes)
+        except Exception as e:  # noqa: BLE001
+            logger.warning(f"frame {frame_idx}: face-service /embed failed: {e}")
+            continue
+
+        if embed_result is None:
+            continue
+
+        embedding_list, _quality = embed_result
+        embedding = np.array(embedding_list, dtype=np.float32)
+
+        new_track = True
+        if last_embedding is not None and last_capture_time is not None:
+            dist = cosine_distance(embedding, last_embedding)
+            age_sec = (captured_at - last_capture_time).total_seconds()
+            if dist < cfg.track_distance_thresh and age_sec < cfg.track_window_sec:
+                new_track = False
+
+        if new_track:
+            track_info = api.create_track(camera_name=cfg.camera_name, first_seen_at=captured_at)
+            current_track_id = track_info["trackId"]
+            current_track_camera_id = track_info["cameraId"]
+            api.add_event(
+                track_id=current_track_id,
+                event_type="arrived",
+                camera_name=cfg.camera_name,
+                occurred_at=captured_at,
+                zone_code=cfg.zone_code,
+            )
+            tracks_created += 1
+            logger.info(f"frame {frame_idx}: new track {current_track_id}")
+
+        assert current_track_id and current_track_camera_id
+
+        try:
+            face.save_track_embedding(
+                jpeg_bytes=jpeg_bytes,
+                track_id=current_track_id,
+                camera_id=current_track_camera_id,
+                captured_at=captured_at,
+            )
+            embeddings_saved += 1
+        except Exception as e:  # noqa: BLE001
+            logger.warning(f"frame {frame_idx}: save_track_embedding failed: {e}")
+
+        last_embedding = embedding
+        last_capture_time = captured_at
+
+    cap.release()
+    logger.info(
+        f"Готово: создано {tracks_created} треков, сохранено {embeddings_saved} эмбеддингов "
+        f"(всего просмотрено {frame_idx} кадров)"
+    )
+    return {
+        "tracks_created": tracks_created,
+        "embeddings_saved": embeddings_saved,
+        "frames_processed": frame_idx,
+    }