# SMAN 1 Garut – Face Embedding Service

Layanan Python kecil untuk menghasilkan **embedding wajah** + metrik kualitas yang dipakai backend CI4 (`FaceService`).

## Fitur

- Framework: **FastAPI** + **InsightFace**.
- Endpoint utama: `POST /embed` (single image → embedding + quality).
- Deteksi wajah, pilih wajah terbesar, hitung:
  - `embedding` (vector float, default dimensi 512),
  - `faces_count`,
  - `face_size` (px),
  - `blur` (variance of Laplacian),
  - `brightness` (0..1),
  - `quality_score` (kombinasi sederhana dari metrik di atas).

## Setup Cepat

```bash
cd face-service
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt

uvicorn main:app --host 0.0.0.0 --port 5000
```

Lalu di `.env` backend CI4 (`backend/.env`), set:

```ini
FACE_SERVICE_URL = 'http://localhost:5000'
FACE_EMBEDDING_DIM = 512
FACE_SIM_THRESHOLD = 0.85
FACE_MIN_SIZE = 80
FACE_MIN_BLUR = 30
FACE_MIN_BRIGHTNESS = 0.2
```

## Kontrak API: `POST /embed`

**Request** (multipart/form-data):

- Field `image`: file gambar (.jpg/.png).

**Response** (`200 OK`, JSON):

```json
{
  "embedding": [0.01, -0.23, ...],
  "quality_score": 0.93,
  "faces_count": 1,
  "face_size": 120.5,
  "blur": 45.2,
  "brightness": 0.55
}
```

Backend CI4 (`FaceService::extractEmbeddingWithQuality`) akan:

- Menolak gambar dengan:
  - `faces_count != 1`,
  - `face_size < FACE_MIN_SIZE`,
  - `blur < FACE_MIN_BLUR`,
  - `brightness < FACE_MIN_BRIGHTNESS`.

## Health Check

`GET /health` → `{"status": "ok", "model": "buffalo_l"}`

Dipakai untuk cek cepat apakah service sudah siap dipakai backend CI4.