Embedding ANPR License Plate Recognition into a Mobile App

A driver parks, the camera scans the plate, but the system outputs "О00Р777" instead of "О 777 РР 777" — a scenario many know well. We face such cases regularly and know how to turn raw OCR output into a reliable result. ANPR (Automatic Number Plate Recognition) on a mobile device is a mature task, but the key engineering challenge is not the OCR itself — it's the pipeline: from frame capture to a validated plate. Night, dirt, reflections, non-standard CIS fonts — each factor requires separate handling. We develop turnkey ANPR solutions for iOS and Android using modern computer vision models. Our multi-year experience in video analytics ensures stable performance even in complex conditions. We'll evaluate your project and suggest the optimal approach — on-device, cloud, or hybrid. Contact us for a preliminary assessment.

How to Achieve Recognition Accuracy in Challenging Conditions?

The choice of approach depends on speed, autonomy, and accuracy requirements. Compare the two main options:

For parking lots and checkpoints with high scan frequency, we recommend on-device. If maximum accuracy is required (e.g., entry verification), we use a cloud fallback.

On-device ANPR: Stack and Example

On iOS, we use Vision + a custom YOLOv8 for plate detection. The model is trained on CIS plate datasets. The code example above includes detection, crop, and OCR.

CIS Plate Normalization

OCR produces raw text. For CIS plates, post-processing is mandatory: replace visually similar characters (0→O, 1→I, 8→B) and validate against regular expressions for each country. We have built a normalizer for RU, BY, UA, KZ. If no pattern matches, we return low confidence.

What If a Plate Fails Validation?

In continuous scanning mode (e.g., checkpoint), we use a frame filter: three consecutive identical results — only then we consider the plate valid. This reduces false positives from random objects resembling a plate. The Android implementation using CameraX is shown in the source code.

Why On-device ANPR Is Better Than Cloud?

On-device processing completes in <50 ms, 4–10 times faster than a cloud request. Full autonomy eliminates network dependency and server costs. Moreover, on-device models (certified solutions) achieve 92–96% accuracy on clean frames, and with cloud fallback up to 97%. This makes on-device ANPR ideal for large-scale high-load deployments.

Typical ANPR Implementation Stages

What's Included

• ANPR module architecture: model selection, pipeline, normalization
• Detection and OCR implementation for iOS and/or Android
• Support for up to 5 countries (extendable)
• Integration with your database via REST API
• Testing on real video streams (night mode, rain, snow)
• Documentation and operation manual
• Post-warranty support

Common ANPR Implementation Mistakes

• Neglecting normalization: OCR sees "0" instead of "O" — result mismatches database
• Too high frame rate: 30 FPS is useless, CPU load increases, accuracy doesn't improve. Optimum is 5 fps
• No noise filtering: one incorrect frame can block entry
• Ignoring regional specifics: Kazakhstan plates consist of digits and Latin letters, while Russian plates use Cyrillic and digits

Timeline Estimates

A basic on-device implementation for one country takes 3–5 days. A full multi-country system with continuous video analysis and integration takes 1–2 weeks. Contact us for an accurate estimate. We have been on the market for over 5 years, completed 30+ video analytics projects — we know how to avoid pitfalls. We guarantee quality at every stage.

More information about ANPR can be found on Wikipedia.

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.