AI Colorization of Black-and-White Photos: Mobile App Integration Guide

How to Integrate AI Colorization for B&W Photos in Mobile Apps

We've often encountered this scenario: a user uploads an old black-and-white photo hoping for natural colors. Photo colorization is a challenging computer vision task. Our focus is AI colorization of black-and-white photos for mobile. But colorization is a problem with one correct input and an infinite number of "correct" outputs. The sky could be blue or gray. The coat could be black or blue. An AI model makes a statistically probable choice, not a restoration of reality. We explain this nuance to the user directly in the UI, but the technical implementation remains the most interesting part. Our experience — 5 years in mobile development and 12+ projects with AI integration — allows us to guarantee a quality result.

Which Model Is Suitable for Mobile Colorization?

The classic choice is DeOldify (fastai + U-Net with self-attention). It generalizes well and produces saturated colors. The downside: it sometimes "pollutes" color with unwanted stains on faces or clothing on complex backgrounds. DDColor (2023) — a transformer-based architecture — works significantly better on portraits and architecture. It more accurately reproduces skin and sky colors. In fact, DDColor is 2x more accurate on skin tones than DeOldify. BigColor and ChromaGAN are for specific tasks (historical photos with severe degradation).

For mobile, we use one of these — converted to Core ML or TFLite. After INT8 quantization, DeOldify weighs 60–70 MB — acceptable for a mobile app.

How to Integrate the Model on iOS and Android?

The integration process consists of several steps:

1. Select the model and convert it to the target format (Core ML / TFLite / ONNX).
2. Configure preprocessing: convert grayscale to RGB, normalize.
3. Run inference with device-specific optimization (Neural Engine, GPU, NNAPI).
4. Post-processing: tiled inference with color normalization for large images.

Converting DeOldify to Core ML

import coremltools as ct
import torch
from deoldify.visualize import get_image_colorizer

colorizer = get_image_colorizer(artistic=True)
model = colorizer.learn.model.eval()

# Export via torch.jit.trace
example = torch.zeros(1, 3, 256, 256)
traced = torch.jit.trace(model, example)

mlmodel = ct.convert(
    traced,
    inputs=[ct.TensorType(name="input", shape=(1, 3, 256, 256))],
    compute_precision=ct.precision.FLOAT16,
    minimum_deployment_target=ct.target.iOS16
)
mlmodel.save("DeOldify_artistic.mlpackage")

An important detail: DeOldify expects RGB input (even for grayscale — it internally converts to Lab and works with AB channels). Before feeding a grayscale image, you must replicate it into a 3-channel RGB: gray → [gray, gray, gray]. In Core ML this is done in preprocessing, not in the model itself.

iOS: Running Inference

let config = MLModelConfiguration()
config.computeUnits = .cpuAndNeuralEngine  // ANE for FLOAT16

let model = try DeOldify_artistic(configuration: config)

// Prepare: grayscale CVPixelBuffer → RGB
func prepareInput(from grayImage: UIImage) -> CVPixelBuffer? {
    // Create RGB pixel buffer from grayscale, replicating channel
    var pixelBuffer: CVPixelBuffer?
    CVPixelBufferCreate(nil, width, height,
                        kCVPixelFormatType_32BGRA, nil, &pixelBuffer)
    // ... copy gray channel into all three
    return pixelBuffer
}

let input = DeOldify_artisticInput(input: pixelBuffer)
let output = try model.prediction(input: input)
let colorizedImage = UIImage(cvPixelBuffer: output.output)

On iPhone 13, a 512×512 image processes in 0.8–1.2 seconds. For Full HD (1920×1080), we use tiled inference with 512×512 patches and blending at seams. The tile-based approach creates a problem of color inconsistency between patches: the model might color one sky fragment blue and a neighboring one gray. The solution is global color histogram matching: normalize each tile's histogram to the global histogram.

Android: TFLite with ONNX Runtime as an Alternative

// ONNX Runtime gives flexibility: one model for iOS and Android
val env = OrtEnvironment.getEnvironment()
val session = env.createSession(
    "deoldify_optimized.onnx",
    OrtSession.SessionOptions().apply {
        addNnapi()  // ONNX Runtime NNAPI delegate for acceleration
    }
)

val inputTensor = OnnxTensor.createTensor(env, inputArray, longArrayOf(1, 3, 512, 512))
val results = session.run(mapOf("input" to inputTensor))
val outputArray = (results[0].value as Array<*>)

ONNX Runtime Mobile is a good choice when the same model is needed on both platforms — no need for double conversion. The NNAPI delegate works on devices with Android 8.1+. ONNX Runtime reduces deployment complexity by 3x compared to separate Core ML and TFLite builds.

UX: What to Consider

Colorization is a slow operation (1–5 seconds). Show an animated progress indicator. A good pattern is a before/after slider: the user drags a divider and sees the original and result side by side. This is both UX and a clear demonstration of the model's work.

Save the result in .heic or .jpg with maximum quality. The colorization result poorly tolerates repeated JPEG compression: artifacts on color transitions become noticeable.

What's Included in the Work

• Selection and conversion of the model for the target platform (Core ML / TFLite / ONNX)
• Optimization for specific devices (Neural Engine, GPU, NNAPI)
• Implementation of tiled inference with color normalization
• UI development with before/after comparison and progress
• Integration of saving to gallery and sharing
• Documentation on the model, preprocessing, and instructions for App Store Review (according to App Store Review Guidelines section 5.1 for photo processing)
• Testing on historical photos of varying quality

Savings on licenses: you don't pay for open-source models. Typical development cost for one platform is $8,000–$12,000, saving you up to $20,000 compared to hiring an in-house team. Development cost is comparable to one developer's monthly salary.

We provide turnkey integration (под ключ) in 2–8 weeks. The work includes (входит) all steps above. Contact us to evaluate your project for free. Write to us for a quote.

Timeline Estimates

We can implement the solution for you in as little as 2 weeks for a single platform. Get a consultation — we will evaluate your project and propose the optimal stack within two days.

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.