boc/quixzoom-capture-pipeline/README.md

QUIXZOOM Capture Pipeline

Datainsamlings- och AI-bearbetningspipeline för QUIXZOOM Field Intelligence Network.

Arkitektur

iPhone Camera
    │
    ▼
Capture App (iOS)
    │
    ▼
Upload API (Node.js/Express)
    │
    ▼
Cloudflare R2 (Object Storage)
    │
    ├── originals/     # Originalbilder
    ├── metadata/      # JSON-metadata per bild
    ├── thumbnails/    # Genererade thumbnails
    └── datasets/      # Träningsdataset
    │
    ▼
AI Preprocessing Worker
    ├── Blur detection
    ├── Duplicate detection
    ├── OCR (Tesseract)
    ├── Object detection (COCO-SSD)
    ├── Scene classification
    ├── Quality scoring
    └── Thumbnail generation
    │
    ▼
Taxonomy Tagger
    ├── Physical layer
    ├── Operational layer
    ├── Economic layer
    ├── Institutional layer
    ├── Social layer
    └── Temporal layer
    │
    ▼
Feature Store / Dataset Builder
    │
    ▼
Training Sets (JSONL-format)

Snabbstart

1. Installation

git clone https://github.com/quixzoom/capture-pipeline.git
cd capture-pipeline
npm install

2. Konfiguration

cp .env.example .env
# Redigera .env med dina värden

3. Starta server

npm start

4. Starta workers

# Terminal 1: AI preprocessing
npm run worker:ai

# Terminal 2: Taxonomy tagger
npm run worker:taxonomy

API Endpoints

Upload

POST /api/upload
Content-Type: multipart/form-data

Form fields:
- image: Bildfil (JPEG/PNG)
- latitude: GPS latitud
- longitude: GPS longitud
- accuracy: GPS noggrannhet (meter)
- altitude: Höjd över havet
- compassHeading: Kompassriktning (grader)
- gyroscope: Gyroskop-data (JSON)
- accelerometer: Accelerometer-data (JSON)
- deviceModel: Telefonmodell
- deviceOS: OS-version
- appVersion: App-version
- contributorId: Zoomer-ID
- missionId: Uppdrags-ID

Status

GET /api/upload/status/:captureId

Dataset

GET /api/dataset/:version
GET /api/datasets

Admin

GET /api/stats
GET /api/queue/status

Metadata Schema

Varje bild sparas med följande metadata:

{
  "captureId": "uuid",
  "uploadedAt": "ISO-8601",
  "processedAt": "ISO-8601",
  "exif": {},
  "gps": {
    "latitude": 0.0,
    "longitude": 0.0,
    "altitude": 0.0,
    "accuracy": 0.0
  },
  "sensors": {
    "compassHeading": 0.0,
    "gyroscope": {},
    "accelerometer": {},
    "deviceOrientation": ""
  },
  "device": {
    "model": "",
    "os": "",
    "appVersion": "",
    "camera": ""
  },
  "environment": {
    "weather": "",
    "temperature": 0.0,
    "sunPosition": "",
    "timeOfDay": ""
  },
  "file": {
    "originalName": "",
    "mimeType": "",
    "size": 0,
    "checksum": "",
    "dimensions": {
      "width": 0,
      "height": 0
    }
  },
  "aiAnalysis": {
    "status": "pending|processing|completed|failed",
    "blurScore": {},
    "qualityScore": {},
    "objectsDetected": [],
    "ocrText": {},
    "sceneClassification": {},
    "duplicates": []
  },
  "taxonomy": {
    "physical": {},
    "operational": {},
    "economic": {},
    "institutional": {},
    "social": {},
    "temporal": {},
    "confidence": 0
  }
}

Taxonomi

Sex-lagers modell baserad på Landvex Urban Intelligence Framework:

Lager	Fråga	Exempel
Physical	Vad finns här?	Byggnader, vägar, fordon, människor
Operational	Vad sker?	Trafik, affärer öppna, konstruktion
Economic	Hur finansieras detta?	Familjeföretag, kedjor, informell ekonomi
Institutional	Vilka regler styr detta?	Zonindelning, skyltar, tillstånd
Social	Vilka nätverk upprätthåller detta?	Turister, migranter, lokala gemenskaper
Temporal	Hur förändras detta över tid?	Rusningstrafik, säsonger, cykler

Miljövariabler

Variabel	Beskrivning	Standardvärde
NODE_ENV	Miljö	development
PORT	Serverport	3000
R2_ENDPOINT	Cloudflare R2 endpoint	-
R2_ACCESS_KEY_ID	R2 access key	-
R2_SECRET_ACCESS_KEY	R2 secret key	-
R2_BUCKET_NAME	R2 bucket name	quixzoom-capture
REDIS_HOST	Redis host	localhost
REDIS_PORT	Redis port	6379
REDIS_PASSWORD	Redis password	-

Licens

MIT © LandveX Inc.