β οΈ BETA VERSION β This app is currently in active development and testing. Features may change and bugs may exist. Report issues β
Features Β· Requirements Β· Getting Started Β· Architecture Β· Privacy
Visual Assist is a native iOS application designed to help visually impaired users navigate their environment safely and independently. Built with Apple's latest frameworks, it leverages the power of:
 LiDAR + ARKit Depth Sensing |
 Vision Text Recognition |
 Core ML Object Detection |
 SwiftUI Modern Interface |
|
|
Visual Assist is currently in beta testing. This means:
|
| Area | Status | Notes |
|---|---|---|
| Navigation Mode | β Working | Core functionality complete |
| Text Reading | β Working | OCR may vary with lighting |
| Object Detection | π Testing | Accuracy improvements ongoing |
| Voice Commands | β Working | English only for now |
| Apple Watch | οΏ½οΏ½ Planned | Coming in future release |
|
Real-time obstacle detection powered by LiDAR sensor technology.
|
Point-and-read OCR with natural speech synthesis.
|
AI-powered scene understanding and description.
|
|
|
|
|
# Clone the repository
git clone https://github.com/yadava5/VisualAssist.git
# Navigate to project
cd VisualAssist
# Open in Xcode
open VisualAssist.xcodeproj| Step | Action |
|---|---|
| 1οΈβ£ | Select your Development Team in Signing & Capabilities |
| 2οΈβ£ | Connect your iPhone Pro via USB |
| 3οΈβ£ | Press β + R to build and run |
Grant permissions β App announces "Visual Assist ready" β Start using!
VisualAssist/
βββ π App/ # Entry point & state
β βββ VisualAssistApp.swift
β βββ AppState.swift
βββ π Views/ # SwiftUI interface
β βββ HomeView.swift
β βββ NavigationModeView.swift
β βββ TextReadingModeView.swift
β βββ ObjectAwarenessModeView.swift
β βββ Components/
βββ π Services/ # Business logic
β βββ LiDARService.swift
β βββ CameraService.swift
β βββ SpeechService.swift
β βββ HapticService.swift
βββ π Models/ # Data structures
βββ π Utilities/ # Helpers
| Framework | Purpose | |
|---|---|---|
| ARKit | LiDAR depth sensing | π΅ |
| Vision | Text recognition (OCR) | π’ |
| Core ML | Object detection | π£ |
| AVFoundation | Camera capture | π |
| Speech | Voice commands | π΄ |
| Core Haptics | Haptic feedback | π‘ |
| Pattern | Usage |
|---|---|
| MVVM | Clean view/logic separation |
| Combine | Reactive @Published properties |
| Swift Concurrency | Modern async/await |
| iOS 26 Design | Liquid glass UI effects |
VisualAssistTests holds 71 tests across the model and utility layers, and
they run on every push.
scripts/resolve-simulator.sh # pick an available iPhone sim
xcodebuild test -project VisualAssist.xcodeproj -scheme VisualAssist \
-destination "platform=iOS Simulator,id=$(scripts/resolve-simulator.sh)" \
-resultBundlePath TestResults.xcresult
scripts/assert-test-results.sh TestResults.xcresult 60Latest run: 71 passed, 0 failed, 0 skipped, verified on iOS 26.5 and again on 26.2.
Two details in that setup are deliberate, and both exist because of mistakes this repository actually made:
- The suite is asserted, not assumed.
xcodebuildexits 0 for a run that executed nothing β no test target attached, a destination matching no tests, a suite skipped wholesale. This project previously shipped a CI check named "Build & Test" that ran no tests at all, green for the life of the branch. The assertion step checks the result bundle for a real pass count, and fails below a floor. - The simulator is resolved, not named. Hard-coding a device model pins CI to one runner image; the images roll forward, the device disappears, and the failure reads like a broken test rather than a missing one.
| Check | Status |
|---|---|
| CI (build + 71 tests) | |
| CodeQL (Swift) |  |
| Secret scanning |  |
| Supply chain | OpenSSF Scorecard |
The Scorecard number is computed and published by the OpenSSF, not by this repository β which is the point of citing it. Several of its 18 checks grade repository settings that no file here can change, so the figure starts modest and the direction of travel matters more than the starting point.
| Feature | Description | |
|---|---|---|
| π | On-Device Processing | All ML runs locally on your iPhone |
| π‘ | No Network Required | Works completely offline |
| π« | No Data Collection | Nothing leaves your device |
| π | No Analytics | Zero tracking or telemetry |
| π€ | No Account | Use immediately, no sign-up |
Visual Assist is built with accessibility as a core principle:
|
|
- β Apple Watch companion app
- πΊοΈ Indoor mapping & saved locations
- π΅ Currency recognition
- π Multi-language support
- π Siri Shortcuts integration
- π CarPlay navigation support
|
This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License.
For commercial licensing, contact the author. |
|
This project uses DocC for API documentation.
# Build documentation in Xcode
# Product β Build Documentation (ββ§βD)
# Or via command line
xcodebuild docbuild -scheme VisualAssist -derivedDataPath ./docsBuilt with β€οΈ for accessibility
Β© 2026 Ayush. All rights reserved.
Visual Assist is not affiliated with Apple Inc.
iPhone, LiDAR, ARKit, and other Apple trademarks are property of Apple Inc.
