feat: add ImageFind application and release pipelines

2026-08-11 18:02:40 +08:00
commit 16239d7525
270 changed files with 59163 additions and 0 deletions
@@ -0,0 +1,119 @@
+#!/usr/bin/env python3
+"""Verify visual/OCR/face GPU indexing and named-person search on live data."""
+
+from __future__ import annotations
+
+import argparse
+import json
+import time
+from pathlib import Path
+from typing import Any
+
+import httpx
+
+
+def call(client: httpx.Client, method: str, path: str, **kwargs: Any) -> httpx.Response:
+    response = client.request(method, path, **kwargs)
+    if response.is_error:
+        raise RuntimeError(f"{method} {path} -> {response.status_code}: {response.text[:1000]}")
+    return response
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--base-url", required=True)
+    parser.add_argument("--token-file", type=Path, required=True)
+    parser.add_argument("--run-dir", type=Path, required=True)
+    parser.add_argument("--case-id", default="face")
+    parser.add_argument("--timeout", type=int, default=600)
+    args = parser.parse_args()
+    state = json.loads((args.run_dir / "state.json").read_text())
+    token = args.token_file.read_text().strip()
+    client = httpx.Client(
+        base_url=args.base_url.rstrip("/"),
+        headers={"Authorization": f"Bearer {token}"},
+        timeout=httpx.Timeout(60, connect=10),
+    )
+    filename = f"{state['run_id']}-{args.case_id}-e2e-face.mp4"
+    observed: dict[str, list[str]] = {name: [] for name in ("visual", "ocr", "faces")}
+    deadline = time.monotonic() + args.timeout
+    video: dict[str, Any] | None = None
+    while time.monotonic() < deadline:
+        models = call(client, "GET", "/api/v1/models").json()
+        components = models.get("accelerator", {}).get("components", {})
+        for component in observed:
+            device = components.get(component, {}).get("actual_device")
+            if device and device not in observed[component]:
+                observed[component].append(device)
+        videos = call(client, "GET", "/api/v1/videos?limit=500").json()
+        video = next(
+            (
+                item
+                for item in videos
+                if item.get("source_id") == state["source_id"]
+                and str(item.get("source_key") or "").rsplit("/", 1)[-1] == filename
+            ),
+            None,
+        )
+        if video and video.get("index_state", {}).get("faces") in {"ready", "failed"}:
+            break
+        time.sleep(1)
+    if not video:
+        raise AssertionError("face fixture never appeared in the video list")
+    if video.get("index_state", {}).get("faces") != "ready":
+        raise AssertionError(f"face indexing failed: {video.get('index_state')} {video.get('error')}")
+
+    matched_person: dict[str, Any] | None = None
+    matched_faces: list[dict[str, Any]] = []
+    for person in call(client, "GET", "/api/v1/people").json():
+        faces = call(client, "GET", f"/api/v1/people/{person['id']}/faces?limit=500").json()
+        owned = [face for face in faces if face.get("video_id") == video["id"]]
+        if owned:
+            matched_person = person
+            matched_faces = owned
+            break
+    if not matched_person:
+        raise AssertionError("the official OpenCV face fixture did not produce a person cluster")
+    person_name = f"{state['run_id']}-人物-729"
+    call(client, "PATCH", f"/api/v1/people/{matched_person['id']}", json={"name": person_name})
+    positive = call(
+        client,
+        "POST",
+        "/api/v1/search",
+        json={"text": person_name, "recognition_types": ["person"], "limit": 100},
+    ).json()
+    if video["id"] not in {item["video_id"] for item in positive.get("items", [])}:
+        raise AssertionError("named-person search omitted the face fixture video")
+    negative = call(
+        client,
+        "POST",
+        "/api/v1/search",
+        json={
+            "text": f"{state['run_id']}-不存在人物-9834721",
+            "recognition_types": ["person"],
+            "limit": 100,
+        },
+    ).json()
+    if video["id"] in {item["video_id"] for item in negative.get("items", [])}:
+        raise AssertionError("non-matching person search returned the face fixture video")
+    non_gpu = {name: devices for name, devices in observed.items() if not any("GPU" in d.upper() for d in devices)}
+    if non_gpu:
+        raise AssertionError(f"GPU was not observed for all frame components: {non_gpu}")
+
+    report = {
+        "video_id": video["id"],
+        "person_id": matched_person["id"],
+        "person_name": person_name,
+        "face_count": len(matched_faces),
+        "observed_devices": observed,
+        "positive_matches": len(positive.get("items", [])),
+        "negative_matches": len(negative.get("items", [])),
+    }
+    (args.run_dir / "face-report.json").write_text(
+        json.dumps(report, ensure_ascii=False, indent=2) + "\n"
+    )
+    print(json.dumps(report, ensure_ascii=False, indent=2))
+
+
+if __name__ == "__main__":
+    main()