Core Problems of Plant Recognition by Photo

A user photographs an unfamiliar flower in the park — within a second expects the name, description, and toxicity warning. Technically, the task is not new: PlantNet and Google Cloud Vision have ready endpoints. But the difference between "works in demo" and "works in production" lies in handling bad photos and UX under low model confidence. Over five years we have accumulated experience in this area and guarantee stable results. The hybrid approach (on-device + cloud) combines offline speed with cloud accuracy, cutting cloud request costs by 2x.

Problems We Solve

• Low-quality photos. Blurry, dark, without leaves or flowers. Without filtering, such images lead to wasted API calls and poor user experience. We use a Laplacian filter to assess sharpness and a separate classifier based on EfficientNet for plant detection.
• Low model confidence. When the top-1 candidate has 30% probability, uncertainty must be presented correctly. We show top-3 with confidence percentages and prompt the user to retake a clearer photo.
• Offline mode. Users expect the app to work without internet, but on-device models lag behind cloud accuracy, especially for rare species. We recommend a hybrid: offline for quick response, cloud for refinement when connection is good.

How We Do It: Stack and Principles

Our core stack: Swift, Kotlin, CoreML/TFLite for on-device, and Plant.id API or PlantNet for cloud verification. We use EfficientNet-B4, fine-tuned on the PlantCLEF dataset, achieving up to 95% accuracy on the test set. For rare species, we employ model ensembles: combining predictions from multiple neural networks boosts overall accuracy by 5–10%. On-device recognition runs 10–30 times faster than cloud, critical for instant feedback.

In one project, we integrated the hybrid approach for a houseplant care app. Users photographed a plant, received instant on-device identification, and with internet access got additional care info (watering, light). This increased retention by 20% and cut cloud API costs by 60%.

Implementation Options: Cloud vs. On-Device

For most projects, a hybrid is optimal: on-device provides a fast hypothesis (top-3), cloud refines when internet is available.

Comparison of Popular APIs

Ensuring Accuracy in Poor Lighting

Before sending to the API, we evaluate image quality: sharpness via Laplacian filter, plant presence via a CoreML classifier. If the image is blurry or lacks a plant, we request a retake, avoiding wasted calls. This reduces cloud API costs by 30–40%. Additionally, we apply auto-exposure and contrast adjustment before submission.

Why Choose the Hybrid Approach?

It combines on-device speed with cloud accuracy. Users see results instantly, and refinement happens in the background. Also, without network, users are not left without an answer. The hybrid architecture reduces cloud computing costs by 2x compared to a pure cloud solution. Our certified specialists (Apple, Google) perform turnkey integration.

Process of Work

1. Analytics — define target audience, usage scenarios, required APIs.
2. Design — choose architecture (cloud/on-device/hybrid), design UX for low confidence.
3. Implementation — integrate APIs, train on-device model, error handling.
4. Testing — validate on a set of 100+ plants, including rare species.
5. Deployment — publish to App Store/Google Play, set up monitoring.

Estimated Timelines

Basic integration of one API (Plant.id or PlantNet) with result handling — 1–2 days. Adding on-device model, quality filtering, offline mode, and cross-platform support — 1–1.5 weeks.

What's Included

• Source code with comments
• API documentation
• Deployment setup (CI/CD)
• Consultations for further customization

Contact us to discuss your project details. Request a consultation on AI recognition integration.

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.