Offline AI on iOS: Using CoreML with SwiftUI

 

📱 Why Offline AI Matters in 2025

With increasing privacy concerns, patchy internet access, and user expectations for instant results, offline AI is becoming a must-have in iOS apps.

CoreML, Apple’s on-device machine learning framework, lets you run AI models directly on the device. That means:

• No network latency
• Full user privacy
• Instant predictions
• Better battery performance
• Works anywhere, anytime

Today, you’ll learn how to integrate CoreML with SwiftUI to build a powerful offline image classifier using a pre-trained ML model.

🎯 Use Case: Offline Image Classifier in SwiftUI

We’ll build an app that:

• Lets users select or take a photo
• Uses MobileNetV2 to classify the image
• Runs entirely offline
• Displays the AI prediction in a sleek SwiftUI UI

🔧 Step 1: Add CoreML Model to Xcode

1. Download MobileNetV2.mlmodel from Apple’s Model Gallery.
2. Drag and drop the file into your Xcode project.
3. Xcode will automatically create a Swift class (MobileNetV2.swift) for easy integration.

⚙️ Step 2: CoreML Prediction Handler

Create a helper class to handle predictions from the model.

import CoreML
import Vision
import UIKit

class MLHandler {
    static let shared = MLHandler()
    private let model = try! VNCoreMLModel(for: MobileNetV2().model)
    func classify(image: UIImage, completion: @escaping (String?) -> Void) {
        let request = VNCoreMLRequest(model: model) { request, _ in
            if let result = request.results?.first as? VNClassificationObservation {
                completion(result.identifier)
            } else {
                completion("Unknown")
            }
        }
        guard let ciImage = CIImage(image: image) else {
            completion(nil)
            return
        }
        let handler = VNImageRequestHandler(ciImage: ciImage)
        try? handler.perform([request])
    }
}

🧠 Step 3: ViewModel to Bridge SwiftUI

import SwiftUI

class ImageClassifierViewModel: ObservableObject {
    @Published var image: UIImage?
    @Published var prediction: String = "No prediction yet"
    func classifyImage() {
        guard let img = image else { return }
        MLHandler.shared.classify(image: img) { [weak self] result in
            DispatchQueue.main.async {
                self?.prediction = result ?? "Error"
            }
        }
    }
}

🎨 Step 4: Build SwiftUI Interface

struct ContentView: View {
    @StateObject var viewModel = ImageClassifierViewModel()
    @State private var showImagePicker = false

var body: some View {
        VStack(spacing: 20) {
            if let image = viewModel.image {
                Image(uiImage: image)
                    .resizable()
                    .scaledToFit()
                    .frame(height: 300)
            }
            Text(viewModel.prediction)
                .font(.headline)
                .padding()
            Button("Select Image") {
                showImagePicker.toggle()
            }
            Button("Classify Image") {
                viewModel.classifyImage()
            }
            .disabled(viewModel.image == nil)
        }
        .sheet(isPresented: $showImagePicker) {
            ImagePicker(image: $viewModel.image)
        }
        .padding()
    }
}

🖼 Image Picker: UIKit Integration in SwiftUI

import SwiftUI
import UIKit

struct ImagePicker: UIViewControllerRepresentable {
    @Binding var image: UIImage?
    func makeUIViewController(context: Context) -> UIImagePickerController {
        let picker = UIImagePickerController()
        picker.delegate = context.coordinator
        return picker
    }
    func updateUIViewController(_ uiViewController: UIImagePickerController, context: Context) {}
    func makeCoordinator() -> Coordinator {
        Coordinator(self)
    }
    class Coordinator: NSObject, UINavigationControllerDelegate, UIImagePickerControllerDelegate {
        let parent: ImagePicker
        init(_ parent: ImagePicker) {
            self.parent = parent
        }
        func imagePickerController(_ picker: UIImagePickerController, didFinishPickingMediaWithInfo info: [UIImagePickerController.InfoKey: Any]) {
            parent.image = info[.originalImage] as? UIImage
            picker.dismiss(animated: true)
        }
    }
}

💡 Why CoreML + SwiftUI Is the Perfect Combo

This integration gives your app:

• 💡 Offline Inference: No network needed
• 🔐 On-device Privacy: No personal data sent to cloud
• ⚡️ Real-time Predictions: Fast, smooth UX
• 🔋 Battery Optimization: CoreML is hardware-accelerated
• 🧩 SwiftUI Friendly: Fully native experience

📢 Final Thoughts

With just a few lines of Swift code, you’ve built an AI-powered iOS app that runs entirely offline — no internet, no servers, just pure machine learning magic.

This is the future of privacy-first, lightning-fast mobile AI. And you’re already building it.