Mobile Image Segmentation Implementation

Mobile Image Segmentation Implementation

We've repeatedly faced the same challenge: a client wants to implement image segmentation in a mobile app but doesn't know where to start. Segmentation is the most computationally expensive computer vision task on mobile. While detection returns a bounding box, segmentation returns a pixel-wise mask. For a 512×512 image, that's 262,144 pixels, each with a class, and all must be processed and rendered within 33 ms for 30 FPS. Without proper model selection and rendering, real-time is unattainable. We've accumulated experience across 10+ computer vision projects and guarantee stable performance.

Semantic or Instance Segmentation?

Semantic segmentation assigns one class per pixel (sky, person, road). All people in frame share the class 'person'. Models: DeepLabV3+, MobileNetV3 Segmentation. The TFLite version of DeepLabV3+ with 257×257 input runs in 22–35 ms on modern Android devices.

Instance segmentation gives each object instance its own mask. Three people = three masks. Models: Mask R-CNN, YOLOv8-seg. Significantly heavier: YOLOv8n-seg on TFLite takes 80–120 ms on mobile. True real-time with such a model requires flagship devices with GPU delegate.

For most consumer use cases (background removal, blur effect), semantic segmentation of 'person' or 'background' is sufficient. This is covered by ML Kit Selfie Segmentation — on-device, 30 FPS, a dedicated neural net trained for this scenario. Our experience shows ML Kit is 2–3× faster than universal models with comparable quality.

How-to: Integrate ML Kit Selfie Segmentation in 5 Days

Integration takes 5–7 days using our checklist:

1. Add SDK via Gradle (Android) or CocoaPods (iOS).
2. Create a SelfieSegmenterOptions instance with STREAM_MODE.
3. Feed frames to segmenter.process() (each camera frame).
4. Obtain the mask and overlay via GPU (Metal/OpenGL).
5. Optimize: use enableRawSizeMask() for quality.

val segmenter = Segmentation.getClient(
    SelfieSegmenterOptions.Builder()
        .setDetectorMode(SelfieSegmenterOptions.STREAM_MODE)
        .enableRawSizeMask()
        .build()
)

Achieving Real-Time Mask Overlay

The segmentation mask is a ByteArray or FloatArray with class indices. Overlaying it onto a video stream within 33 ms requires GPU.

On iOS, we use Metal for blending: the mask is converted to CIImage via CIFilter.pixellate or a custom Metal kernel, then blended with the original frame using CIBlendWithMask. All Metal rendering avoids CPU-GPU copy overhead via MTLBuffer with shared storage mode.

On Android, we use RenderScript (deprecated in API 31+) or Vulkan/OpenGL ES via SurfaceView. For newer projects: AGSL (Android Graphics Shading Language) starting Android 13, or Canvas.drawBitmap with Paint.xfermode = PorterDuffXfermode(PorterDuff.Mode.DST_IN) for simple cases.

A common mistake: generating a Bitmap from the mask on CPU for every frame. On Pixel 6 this takes ~18 ms just for allocation and copy — eating the entire 33 ms budget. The correct approach: use Bitmap.copyPixelsFromBuffer with a pre-allocated ByteBuffer or pass the mask directly to a shader.

ML Kit Selfie Segmentation in Practice

The fastest path for “background blur” in video calls or photo editors is ML Kit. Case study: a corporate video presentation app with virtual background on iOS. Core Image CIBlendWithMask + ML Kit Selfie Segmentation (iOS SDK): 28 ms on iPhone 13 mini at 720p. On iPhone SE 2nd gen — 41 ms, causing dropped frames every 2–3 seconds at 30 FPS. Solution: reduce processing resolution to 540p, upscale mask using bilinear interpolation — 24 ms, dropped frames gone.

What Our Work Includes

• Requirement analysis and optimal model selection (ML Kit / custom / OpenCV).
• Pipeline design (frame capture → segmentation → overlay → rendering) accounting for target FPS.
• Implementation: coding in Swift/Kotlin, integrating TFLite/Metal/OpenGL.
• Testing on real devices (from iPhone SE to flagship Android).
• Task-specific optimization: profiling, latency reduction.
• Integration documentation and training for the client's team.

Timeline and Cost

ML Kit Selfie Segmentation integration with effect overlay — 5–7 days. Custom segmentation model with Metal/OpenGL rendering on real-time video — 2–3 weeks. Cost is determined individually after analysis. We have over 5 years of market experience and 15+ computer vision projects completed. Order turnkey segmentation development and get a ready solution. Get a consultation for your project.

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.