OCR text recognition implementation in mobile apps

Receipts with faded ink, crumpled invoices, handwritten notes — standard OCR libraries often fail on such documents. We solve this with custom preprocessing and post-processing: we choose the stack (Vision, ML Kit, or Tesseract) for each specific task rather than using a one-size-fits-all solution. Get a free project assessment — contact us.

How to choose the right OCR tool?

Choosing between Vision, ML Kit, and Tesseract depends on your task and requirements for speed and accuracy. iOS Vision framework — VNRecognizeTextRequest — works on-device, supports 18+ languages including Cyrillic. The .accurate level takes 180–350 ms to process an A4 photo on an iPhone 12. ML Kit Text Recognition v2 is cross-platform, on-device, supports Cyrillic and CJK. On Android, use TextRecognition.getClient(TextRecognizerOptions.DEFAULT_OPTIONS). Tesseract via SwiftyTesseract / tess-two is for custom fonts, but 3–5 times slower. For example, ML Kit on Android is 1.5 times faster than Tesseract on the same device for printed text. We help you pick the optimal tool for your task — from simple scanning to table and form recognition.

Why preprocessing is critical?

40% of OCR success comes from input image quality. A typical pipeline:

1. Grayscale — remove color noise
2. Contrast correction via CIColorControls (iOS) or ColorMatrix (Android)
3. Binarization (Otsu threshold) — for uneven lighting
4. Deskew — correct perspective and tilt

From our practice: for an invoice scanning app, preprocessing boosted accuracy from 78% to 94% in a warehouse setting. Perspective correction using VNDetectRectanglesRequest and CIPerspectiveCorrection was especially effective. Practical results:

How we implement OCR: a real case

We built OCR for a logistics client: recognizing invoice numbers with dot-matrix font on crumpled boxes. We chose ML Kit with custom preprocessing. Pipeline: grayscale → Otsu binarization → deskew → ML Kit. Post-processing: extract invoice number via regex with checksum. Accuracy on test set: 96%. Timeline: 5 business days.

func preprocessImage(_ image: UIImage) -> UIImage {
    guard let ciImage = CIImage(image: image) else { return image }
    let gray = ciImage.applyingFilter("CIPhotoEffectMono")
    let contrast = gray.applyingFilter("CIColorControls", parameters: [kCIInputContrastKey: 1.1])
    // Otsu binarization not built-in, use custom kernel
    return UIImage(ciImage: contrast)
}

This pipeline is standard in every project. We guarantee results.

What does post-processing provide?

Recognized text is not always ready data. For extracting phone numbers, emails, dates, we use NSDataDetector (iOS) or Patterns (Android). For structured fields (INN, SNILS) — regex with checksum validation. Importantly, ML Kit returns TextBlock → TextLine → TextElement with coordinates. We group lines by Y to reconstruct the table structure. This reduces manual entry by 30–50%.

Tool comparison

How to recognize text in real time?

For real-time OCR, we reduce resolution to 720p and run recognition every 3-5 frames with result buffering. On Android, ML Kit in STREAM_MODE manages frequency automatically. On iOS — AVCaptureVideoDataOutput with VNRecognizeTextRequest. This gives a stable 15-20 fps without overheating. Order such a module — get integration consultancy.

How we work on an OCR module

1. Analysis: study document types, shooting conditions, fonts.
2. Tool selection and preprocessing pipeline design.
3. Recognition and post-processing implementation.
4. App integration (batch or real-time).
5. Testing on a set of real data (at least 100 samples).
6. Deployment and support.

What's included in OCR module development

• Requirements analysis and stack selection
• Preprocessing pipeline development for your conditions
• Recognition implementation (photo or real-time)
• Data post-processing (field extraction, validation)
• App integration (iOS/Android)
• Testing and refinement based on results
• Documentation and code review
• Our team's experience: over 5 years in mobile, 30+ OCR projects

Timelines and cost

Timelines: from 3 business days (photo OCR with preprocessing) to 2 weeks (real-time document scanner with correction). OCR implementation reduces manual entry costs by 30–50%. Cost is calculated individually — contact us for an estimate. We guarantee quality: you receive a ready module with at least 95% accuracy (on agreed document types). Order development or get a consultation.

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.