ML Kit Integration for Mobile Apps: Custom & Ready-Made APIs

ML Kit for Mobile Apps: Integration and Optimization

We integrate Google ML Kit into mobile apps for text recognition, face detection, object detection, and other AI features. In practice, clients often come with a task like: 'We need a receipt scanner that works on Android and iOS with >95% accuracy.' Behind the seeming simplicity of ML Kit lie nuances that only surface in production—from image orientation to speed differences on budget devices. We've gathered real implementation experience so you don't step on the same rake.

Common Problems When Working with ML Kit

Ready-made APIs (Text Recognition v2, Face Detection, Barcode Scanning) work correctly if the input image requirements are met. Face Detection with FaceDetectorOptions.PerformanceMode.ACCURATE on Android returns results in 80–150 ms on a Pixel 6, but on budget devices with Snapdragon 680 it's already 400+ ms. Using FAST mode drops accuracy when the head is rotated more than 30°.

On iOS, MLKitFaceDetection via VisionImage(image:) loses image orientation if image.orientation is not explicitly set from UIImage.imageOrientation. No crash occurs—faces simply aren't detected when the phone is held horizontally.

With custom TFLite models via CustomImageLabeler, metadata packaging is crucial. Without TFLiteMetadataHelper, the model doesn't know input normalization—you must either add metadata via flatbuffers or specify normalization manually through CustomRemoteModel options.

Why On-Device Often Beats Cloud?

On-device works offline, is faster, and incurs no API call costs. Cloud is more accurate for complex cases (multilingual OCR, non-standard fonts). For most B2C apps, a hybrid scheme is optimal: on-device as the primary path, cloud as fallback when confidence is low. This way we achieved 94% accuracy on standard receipts without paying for cloud calls.

How We Implement ML Kit: Receipt Recognition Case Study

A case from practice: an app for scanning receipts. ML Kit Text Recognition v2 on-device gave 94% accuracy on standard cash receipts, but only 67% on faded thermal paper. We added preprocessing via CIFilter (increased contrast, binarization) before passing to VisionImage—accuracy rose to 89% without switching to Cloud API.

For Android, integration goes through BarcodeScanning.getClient() or TextRecognition.getClient(TextRecognizerOptions.DEFAULT_OPTIONS). Models are auto-downloaded via Play Services on first launch—this must be accounted for in UX: the first inference may take several seconds until the model is loaded. We use ModuleInstallClient for explicit preloading during onboarding.

For custom models—FirebaseModelDownloader with ModelDownloadType.LOCAL_MODEL_UPDATE_IN_BACKGROUND. The model updates in the background; the app uses the current version until the next launch.

How to Avoid Common Integration Mistakes?

• Always check image.orientation on iOS—otherwise face detection breaks when the phone orientation changes.
• On Android, use ModuleInstallClient to preload models and avoid first-inference delay.
• For custom TFLite models, always pack metadata via TFLiteMetadataHelper; otherwise normalization will be wrong.
• Test on 5-10 real devices from different manufacturers—inference speed can vary by 3x.
• Implement fallback logic: when on-device confidence is low, send the request to Cloud API (if the task is critical).

Supported ML Kit APIs

What's Included in ML Kit Integration

We deliver a complete package:

• Audit of current architecture and selection of optimal API (ready-made vs custom)
• Integration of ML Kit SDK with image preprocessing configuration
• Testing on 10+ devices (different OS versions, manufacturers)
• Integration and configuration documentation
• Team training on working with models
• 6-month warranty after delivery

Process and Timeline

Requirements audit → API selection (ready-made vs custom) → SDK integration → preprocessing setup → testing on target devices → production accuracy monitoring setup.

Integration of one ready-made API (e.g., Barcode Scanning or Face Detection) — 2–4 business days. Custom TFLite model with preprocessing and fallback logic — 1–2 weeks. Pricing is determined individually.

We have implemented ML Kit in 15+ projects, including fintech and retail, and guarantee accuracy of at least 90% at acceptance stage. Contact us to evaluate your project—we'll find the optimal solution for your needs.

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.