Mobile App for Gardeners: Offline Plant Recognition & Smart Watering

Mobile App for Gardeners: Smart Plant Recognition & Offline Watering Planner

Picture this: a gardener uploads a photo of a suspicious spot on a tomato leaf — in seconds the app returns a diagnosis (late blight, 87% confidence) and recommends copper sulfate treatment. All offline, in a remote countryside. Building this for iOS and Android requires attention to every detail: from ML model selection to background synchronization.

We create apps where CV models, weather triggers, and offline mode work as a single mechanism. We bring 6 years of mobile development and 50+ AgriTech projects. The challenge isn't individual components, but their seamless integration: rain sensor via OpenWeatherMap, local SQLite database, recognition via CoreML or TFLite — all running reliably on five-year-old devices.

How Plant Recognition Works

The core feature — identifying plants and diseases from photos. Two approaches: cloud API (Plant.id, PlantNet) or on-device model (CoreML/TFLite). Let's compare:

On-device recognition is 5x faster than cloud — 0.2 s vs 1–3 s — but accuracy is 15–20% lower. For a cabin without internet, it's the only option.

The Plant.id API returns the name, diseases with confidence score, and treatment recommendations. The photo is base64-encoded and sent via POST; the response contains suggestions with probability. Important: the API requires a well-lit shot of a leaf or flower — a wide shot gives low accuracy. We always teach the user proper shooting technique.

struct PlantIdentificationRequest: Encodable {
    let images: [String]       // base64
    let modifiers: [String]    // ["crops_fast", "similar_images"]
    let plant_language: String // "ru"
    let plant_details: [String] // ["common_names", "url", "description", "treatment"]
}

The on-device approach uses models from iNaturalist or trained on the PlantVillage dataset (54,000 images, 38 leaf disease classes). Accuracy is 15–20% lower than cloud, but fully offline.

Why Weather Integration Matters

Sounds simple, but a common mistake is setting fixed-time notifications and wondering why users miss waterings. The problem: notifications aren't recalculated when precipitation changes.

The correct logic: every morning fetch the weather forecast via the OpenWeatherMap API or Apple WeatherKit. If rain >5 mm is forecasted, skip watering and cancel the notification via UNUserNotificationCenter.removePendingNotificationRequests. This requires a background task: BGAppRefreshTask on iOS or WorkManager on Android.

On Android we use WorkManager with PeriodicWorkRequest and NetworkType.CONNECTED constraint. Not AlarmManager directly — on Android 12+ SCHEDULE_EXACT_ALARM permission is required and rarely granted by users.

Time saved on manual watering planning — up to 2 hours per week. Water savings from smart rainfall accounting — up to 30% per season.

What's Included in the Work

Ordering a turnkey development includes:

• Architecture design and stack selection (iOS/Android/cross-platform)
• Integration of all APIs (weather, recognition)
• Offline plant database with image caching
• Configuration of push notifications and background tasks
• Operations documentation and access links
• Assistance with publishing to App Store and Google Play
• Code warranty — 3 months of free support

Offline Mode & Plant Database

The local plant database (name, description, care instructions, sowing calendar) is stored in SQLite. For 500–1000 records we use Room on Android, Core Data or GRDB on iOS. Images are cached on first viewing with an LRU policy (Kingfisher on iOS, Coil on Android).

No internet at the dacha — a reality. All basic functions (adding plants, viewing tips, setting reminders) work offline. Synchronization when connection is restored runs via a queue of deferred operations.

Weather Integration

OpenWeatherMap is the standard for such apps: the free tier covers 1000 requests per day. WeatherKit on iOS (from newer versions) is more accurate and doesn't require your own key, but is only available on Apple platforms. For a gardening app, besides temperature, humidity, uvi, and rain (1-hour and 3-hour precipitation) are crucial — available in OWM's current endpoint.

Process

1. Analytics — define the feature set: which plants (only vegetables or also garden + houseplants), whether a social component is needed, whether an offline disease database is required.
2. Design — create the architecture, choose the stack (native or cross-platform).
3. Development — offline database → plant addition → watering schedule with notifications → weather integration → recognition. Recognition last because API keys and pricing need agreement.
4. Testing — cover core logic with unit tests, run field tests on real devices.
5. Deployment — publish to stores, set up Crashlytics monitoring and analytics.

Contact us for a consultation on your project — we'll help choose the optimal stack and estimate deadlines.

Timeline Guide

To estimate your project, contact us — we'll prepare a precise quote and suggest the best solution. We guarantee passing App Store Review and Google Play Review on the first attempt.

Machine Learning in Mobile Apps: CoreML, TFLite, and On-Device Models

We distinguish two fundamentally different approaches: an app with on-device AI and an app that simply calls a cloud API. The former works without internet, does not send user data to third-party servers, and responds within 50 milliseconds. The latter depends on network latency and pricing plans. Choosing the architecture is a key step that directly affects cost, privacy, and user experience in machine learning in mobile apps. Our experience shows that in 70% of projects, on-device inference is cheaper in the long run due to eliminating server costs.

How to Choose Between CoreML and TFLite for On-Device Inference?

CoreML — Apple's native framework for running ML models on device. Supports Neural Engine (starting with A11 Bionic), GPU, and CPU as fallback. Models are converted to .mlmodel format via coremltools from PyTorch, ONNX, or TensorFlow. Conversion is not always trivial: custom layers require implementing MLCustomLayer, and INT8 quantization can sometimes noticeably reduce accuracy on specific data. We ensure the final model passes validation on real data before and after conversion.

TensorFlow Lite — cross-platform alternative for Android and Flutter. On Android it uses NNAPI (Neural Networks API) for hardware acceleration — since Android 10 NNAPI is more stable; before that it's better to explicitly use GPU delegate via GpuDelegate. A typical mistake: the model is trained on normalized data in range [0,1], but the app feeds [0,255] — inference runs but produces meaningless results without any error. We include an automatic input data validation module in the SDK.

For image classification, object detection, and segmentation tasks, ready-to-use optimized models are available. YOLOv8 in CoreML format runs detection on a 640×640 frame in 15–20 ms on iPhone 14 Neural Engine. MobileNetV3 on TFLite with GPU delegate runs around 8 ms on Pixel 7 for classification.

What If You Need Text Generation On-Device?

Running small language models on device has become a reality in the last few years. Apple Intelligence uses its own models via Private Cloud Compute, but for third-party developers other paths are available.

llama.cpp with Metal backend on iOS is a working approach for phi-3-mini (3.8B parameters, 4-bit quantization, ~2.3 GB). Inference: 15–25 tokens/second on iPhone 15 Pro. For integration in Swift, use the Swift Package llama.swift or a wrapper via C interface llama.h. The binary is not bundled with the app — the model is downloaded on first launch and stored in Application Support. Our certified developers configure incremental download to avoid blocking the first launch.

On Android, the analog is Google AI Edge (formerly MediaPipe LLM Inference API) supporting Gemma-2B. It works via GPU delegate, on Tensor G3 chip Pixel 8 Pro — about 20 tokens/second.

Limitations are real: models larger than 4B parameters are still slow on mobile devices. For complex reasoning tasks, on-device LLM falls behind GPT-4o in quality. A hybrid approach — on-device for short tasks and private data, cloud for complex queries — is often optimal. We will evaluate your case and propose a balance of performance and privacy — contact us.

How Does On-Device Inference Compare to Cloud in Terms of Cost and Performance?

On-device inference is typically 10x cheaper per request than cloud APIs for image recognition tasks, while also eliminating latency variability and privacy risks. The table below summarizes the trade-offs.

For an app with 100k MAU running 10 image recognitions per user per month, on-device inference can save up to $5,000 monthly compared to cloud API. Get a free consultation on your ML architecture today.

Integrating OpenAI API and Other Cloud Models

For scenarios where cloud inference is acceptable, integrating OpenAI, Anthropic, or Google Gemini is an HTTP client + streaming SSE. In Swift, AsyncThrowingStream is convenient for streaming responses. In Kotlin, use Flow.

Critically: API keys must never be stored in the app bundle. Even an obfuscated key can be extracted from the IPA in 10 minutes using strings or frida. Correct architecture: mobile app → your own backend → OpenAI API. The backend controls rate limiting, logs requests, and protects the key.

What Is Included in the Work (Deliverables)

• Trained and quantized model for the target device (documentation with metrics)
• SDK for integration (Swift/Kotlin/Flutter) with call examples
• Performance tests on 3–5 real devices
• Instructions for OTA model updates
• Support during App Store / Google Play moderation (compliance with Guidelines 4.2, 5.1)
• 2 weeks of technical support after release

Typical Project Pipeline

1. Task analysis — measure latency, privacy, size, supported devices.
2. Model prototyping — in Python, evaluate accuracy on target data.
3. Conversion and quantization — for CoreML/TFLite with validation.
4. Integration into the app — model wrapped in a service layer (easy to swap CoreML ↔ TFLite ↔ cloud).
5. Testing — on real devices, measure FPS, RAM, battery.
6. Deployment — via TestFlight / Firebase App Distribution, monitor metrics.

Timelines: integration of a ready CoreML/TFLite model — 1–2 weeks, development of a custom model with mobile optimization — from 6 weeks, on-device LLM chat with personalization — 4–8 weeks.

Why We Take on Complex Cases?

10+ years of experience in mobile development, 50+ implemented AI/ML solutions, guarantee of compatibility with current iOS and Android versions. All projects undergo code review and load testing. The cost includes preparation of moderation documentation and training of your team.

Contact us — we will help you choose the architecture and implement ML in your app turnkey. Order an audit of your existing solution — we will assess the potential for server cost savings free of charge. In some projects, savings can reach significant amounts per month.