FarmVision

AI-powered fruit detection and agricultural analysis platform. Detects fruit on trees from drone and field imagery using YOLO-based models, estimates yield, and generates field health reports from multispectral orthophotos.

Features

• Fruit detection (mandarin, apple, pear, peach, pomegranate, and tree counting) with confidence scoring
• Batch multi-image detection
• Drone orthophoto processing with vegetation index analysis (NDVI, EVI, SAVI and 20+ others)
• Yield prediction and decision engine recommendations
• Async task processing via Celery
• REST API with OpenAPI documentation
• System monitoring dashboard
• Comprehensive test suite

Requirements

• Docker 24+ and Docker Compose v2
• NVIDIA GPU with CUDA 11.8+ (required for inference)
• NVIDIA Container Toolkit installed on host
• 8GB+ GPU VRAM recommended
• SSL certificate (for production)

Quick Start (Docker)

1. Clone and configure environment

git clone https://github.com/ChronoCoders/farmvision.git
cd farmvision
cp .env.example .env

Edit .env and fill in all required values. At minimum: DJANGO_SECRET_KEY=<generate with: python -c "from django.core.management.utils import get_random_secret_key; print(get_random_secret_key())"> DATABASE_PASSWORD=

2. Generate SSL certificate

For development (self-signed):

bash scripts/generate-ssl.sh

For production, place your certificate files at: ssl/cert.pem ssl/key.pem

3. Start services

# Without nginx (direct gunicorn access on port 8000)
docker compose up -d

# With nginx (recommended for production, ports 80/443)
docker compose --profile with-nginx up -d

4. Create superuser

docker compose exec web python manage.py createsuperuser

5. Verify http://localhost:8000/health/ → {"status": "ok", "database": "connected"} http://localhost:8000/admin/ → Django admin http://localhost:8000/docs/ → Swagger UI

Development & Testing

Running Tests

The project includes a comprehensive test suite. To run all tests without external dependencies (Redis/Celery mocked):

# Using Django test runner
python manage.py test --settings=yolowebapp2.test_settings

# Using pytest (recommended)
pytest

Local Setup (Without Docker)

1. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # Windows: .\venv\Scripts\activate

2. Install dependencies:

   pip install -r requirements.txt

3. Run migrations:

   python manage.py migrate

4. Start development server:

   python manage.py runserver

Environment Variables

All variables are read from .env in the project root. See .env.example for the full template.

Variable	Required	Default	Description
DJANGO_SECRET_KEY	Yes	—	Django secret key. Generate with get_random_secret_key().
DJANGO_ENVIRONMENT	Yes	production	production or development
DJANGO_DEBUG	No	False	Set True only in development.
DJANGO_ALLOWED_HOSTS	Yes	localhost	Comma-separated list of allowed hostnames.
DATABASE_NAME	No	farmvision	PostgreSQL database name.
DATABASE_USER	No	farmvision_user	PostgreSQL username.
DATABASE_PASSWORD	Yes	—	PostgreSQL password.
DATABASE_HOST	No	db	Database host (Docker service name).
DATABASE_PORT	No	5432	Database port.
CELERY_BROKER_URL	No	redis://redis:6379/0	Celery broker URL.
CELERY_RESULT_BACKEND	No	redis://redis:6379/0	Celery result backend URL.
REDIS_CACHE_URL	No	redis://redis:6379/1	Redis cache URL (separate DB from broker).
REDIS_PASSWORD	No	—	Redis password. Leave empty if Redis has no auth.
CORS_ALLOWED_ORIGINS	No	—	Comma-separated CORS origins, e.g. https://app.example.com
ALERT_WEBHOOK_URL	No	—	Slack/Teams/Discord webhook for model degradation alerts.
ALERT_EMAIL_RECIPIENTS	No	—	Comma-separated email addresses for degradation alerts.
ALERT_COOLDOWN_SECONDS	No	3600	Minimum time between duplicate alerts.

Services

Service	Port	Description
web	8000	Django + Gunicorn
db	5432 (localhost only)	PostgreSQL 15 with PostGIS
redis	6379 (localhost only)	Redis 7 (broker + cache)
celery_worker	—	Celery task worker
celery_beat	—	Celery periodic task scheduler
nginx	80, 443	Reverse proxy (profile: with-nginx)

YOLO Models

Place model weight files in the models/ directory before starting: models/ ├── mandalina.pt ├── elma.pt ├── armut.pt ├── seftale.pt └── nar.pt

The application will not perform inference if the corresponding .pt file is missing.

API Endpoints

Full interactive documentation available at /docs/ (Swagger) and /redoc/ (ReDoc) after startup.

Detection

Method	Endpoint	Description
GET	/api/detections/	List detection results
POST	/api/detections/	Create detection result
GET	/api/detections/statistics/	Aggregate statistics
GET	/api/detections/recent/	Last 10 results
GET	/api/batches/	List batch detections

Web Endpoints (session auth)

Method	Endpoint	Description
GET/POST	/detection/	Single image detection
GET/POST	/detection/mcti/	Multi-image batch detection
POST	/detection/async-detection/	Async detection (returns task ID)
GET	/detection/task-status/<task_id>/	Check async task status
GET	/detection/system-monitoring/	System resource dashboard
POST	/detection/cache/invalidate/	Invalidate prediction cache (admin only)
GET	/detection/cache/statistics/	Cache hit/miss statistics (admin only)

Drone Mapping

Method	Endpoint	Description
GET	/dron-map/projects/	Project list
GET/POST	/dron-map/projects/<slug>/	Create/edit project
GET	/dron-map/map/<id>/	Interactive map view
GET	/api/projects/	REST API project list
GET	/api/projects/statistics/	Project statistics

System

Method	Endpoint	Description
GET	/health/	Health check (no auth required)
GET	/api/schema/	OpenAPI schema
GET	/docs/	Swagger UI
GET	/redoc/	ReDoc UI

Celery Tasks

Periodic tasks run automatically via celery_beat. Manual execution:

# Run model health check immediately
docker compose exec celery_worker celery -A yolowebapp2 call detection.tasks.check_model_health

# Run cleanup immediately (removes results older than 30 days)
docker compose exec celery_worker celery -A yolowebapp2 call detection.tasks.cleanup_old_results --kwargs='{"days_old": 30}'

# Monitor active tasks
docker compose exec celery_worker celery -A yolowebapp2 inspect active

Task	Schedule	Description
check_model_health	Every 24h	Checks model confidence degradation. Sends alerts if below threshold (0.7).
cleanup_old_results	Every 24h	Deletes detection records and media files older than 30 days.

Running Tests

# Run all tests
docker compose exec web python manage.py test detection dron_map

# Run with pytest (includes yield_prediction and decision_engine tests)
docker compose exec web pytest

# Run specific app
docker compose exec web python manage.py test detection

# With coverage
docker compose exec web coverage run manage.py test && coverage report

Development Setup (without Docker)

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -r requirements.txt

# Use SQLite for local dev (no DATABASE_* env vars needed)
export DJANGO_ENVIRONMENT=development
export DJANGO_SECRET_KEY=dev-secret-key-not-for-production

python manage.py migrate
python manage.py createsuperuser
python manage.py runserver

For async tasks in development, Redis is still required:

# Start Redis locally
docker run -d -p 6379:6379 redis:7-alpine

# Start Celery worker
celery -A yolowebapp2 worker -l info

Logs

# Application logs
docker compose logs web
docker compose logs celery_worker

# Follow in real time
docker compose logs -f web

# Log files inside container
docker compose exec web tail -f /app/logs/django.log

Backup

# Database backup
docker compose exec db pg_dump -U farmvision_user farmvision > backup_$(date +%Y%m%d).sql

# Restore
docker compose exec -T db psql -U farmvision_user farmvision < backup_20240101.sql

Updating

git pull origin main
docker compose build
docker compose up -d
docker compose exec web python manage.py migrate

Security Notes

• Never commit .env to version control
• DATABASE_PASSWORD and REDIS_PASSWORD are required in production
• SSL must be enabled in production (see nginx.conf)
• The /detection/cache/invalidate/ endpoint requires Django staff (is_staff=True) privileges
• Media files are served through authenticated proxy — direct /media/ access requires login
• Rate limiting is enforced at nginx level (10 req/s for API, 100 req/s general)

License

Proprietary. All rights reserved.