VisionLearn

Accessible learning for children with Cortical Visual Impairment (CVI)

Motivation and Background

VisionLearn is a cross-platform mobile application (iOS & Android) designed to support accessible learning experiences for children with Cortical/Cerebral Visual Impairment (CVI).

What is CVI?

Cortical Visual Impairment (CVI) is the leading cause of visual impairment in children in developed countries. Unlike ocular blindness, CVI affects how the brain processes visual information. Children with CVI often have normal eye exams but struggle with visual recognition, attention, and processing.

The Problem

Parents, teachers, and therapists working with CVI children need specialized learning tools that:

Adapt to individual visual profiles
Follow evidence-based protocols like the Perkins CVI framework
Provide high contrast, simplified visuals
Offer audio feedback and accessibility features
Track progress for IEP meetings

Existing apps don't address these specific needs, forcing educators to manually create materials. VisionLearn solves this by providing AI-powered, adaptive learning activities that follow the Perkins CVI Protocol.

Features

Visual Profile System

Based on the Perkins 16 Visual Behaviors framework
Customizable CVI phase selection (Phase I, II, III)
Personalized color preferences (yellow, red, pink, blue, green, orange)
Adjustable complexity and timing settings

Five Learning Modules

Image Recognition - Identify objects with audio feedback and encouragement
Cause & Effect - Touch-response activities for engagement and learning
Sorting - Group objects by category with visual feedback
Matching - Find matching pairs with progressive difficulty
Sequencing - Arrange items in order (numbers, patterns, sizes)

Accessibility First

Large touch targets (minimum 64dp)
High contrast CVI-friendly color options
Screen reader support (TalkBack/VoiceOver)
Audio feedback for all interactions
Text-to-Speech for instructions and encouragement
Configurable response timeouts
Simple, uncluttered visual design

AI-Powered Learning (Google Gemini)

Dynamic question generation tailored to CVI needs
Adaptive difficulty based on performance
Personalized encouragement messages
Fallback to local content when offline

Custom Activity Creator

Create personalized activities with your own images
Take photos or select from gallery
Configure question text and correct answers
Share activities across sessions

Progress Tracking

Session history and statistics
Accuracy tracking per module
Total activities completed
Visual progress dashboard

Screenshots

Onboarding Flow

Welcome	Child's Name	CVI Phase	Color Selection

Home & AI Generation

Home Screen	AI Session Generation

Learning Modules

Image Recognition	Cause & Effect	Sorting

Matching	Sequencing

Progress & Tools

Progress Tracking	Content Creator	Visual Profile

iOS Version

iOS AI Loading

Demo Video

Watch Demo Video

Technical Implementation

Architecture

Component	Technology
UI	Compose Multiplatform 1.7.3
Language	Kotlin 2.1.0
Architecture	MVVM + Clean Architecture
Navigation	Voyager 1.1.0-beta02
Database	SQLDelight 2.0.2
Networking	Ktor 3.0.3
DI	Koin 4.0.0
Image Loading	Coil 3.0.4
AI	Google Gemini API

Shared Code Structure

The app follows Clean Architecture principles with maximum code sharing:

Domain Layer - Business models, repository interfaces
Data Layer - Repository implementations, database, API clients
Presentation Layer - ViewModels (ScreenModels), UI Components
AI Layer - Gemini integration with local fallback

Platform-Specific Code

Feature	Android	iOS
Database Driver	Android SQLite Driver	Native SQLite Driver
Text-to-Speech	Android TTS	AVSpeechSynthesizer
Image Picker	ActivityResultContracts	PHPickerViewController
Accessibility	TalkBack	VoiceOver

AI Integration

VisionLearn uses Google Gemini API for intelligent content generation:

GeminiSessionGenerator - Creates dynamic, CVI-appropriate questions
LocalSessionGenerator - Offline fallback with pre-defined content
Automatic fallback when API is unavailable
Timeout handling for responsive UX

How to Run

Prerequisites

Android Studio Hedgehog (2023.1.1) or later
Xcode 15+ (for iOS)
JDK 17
KDoctor for environment verification

# Verify your environment
brew install kdoctor
kdoctor

1. Clone the Repository

git clone https://github.com/arulagarwal/VisionLearn.git
cd VisionLearn

2. Configure Gemini API Key (Optional but Recommended)

Get a free API key from Google AI Studio:

Create/open local.properties in the project root
Add your API key:

GEMINI_API_KEY=your_api_key_here

Note: The app works without an API key using local fallback content, but AI-generated questions provide a better experience.

3. Build & Run

Android

# Build debug APK
./gradlew :composeApp:assembleDebug

# Install on connected device/emulator
./gradlew :composeApp:installDebug

Or open in Android Studio and run the composeApp configuration.

iOS

# First, build the Kotlin framework
./gradlew :composeApp:linkDebugFrameworkIosSimulatorArm64

# Open Xcode project
open iosApp/iosApp.xcodeproj

In Xcode:

Select an iPhone Simulator (iPhone 15 recommended)
Press Cmd+R to build and run

Note: First iOS build takes 2-5 minutes to compile the Kotlin framework.

Troubleshooting

Issue	Solution
"No such module 'ComposeApp'" in Xcode	Run `./gradlew :composeApp:linkDebugFrameworkIosSimulatorArm64` first
Gradle build fails with Java error	Ensure `JAVA_HOME` points to JDK 17
iOS linker errors for sqlite3	Clean build in Xcode (Cmd+Shift+K) and rebuild

Project Structure

VisionLearn/
├── composeApp/
│   └── src/
│       ├── commonMain/              # Shared Kotlin code (95%+)
│       │   ├── kotlin/
│       │   │   ├── ai/              # AI services (Gemini + Local)
│       │   │   ├── data/            # Repositories, database
│       │   │   ├── di/              # Koin dependency injection
│       │   │   ├── domain/          # Models, interfaces
│       │   │   ├── platform/        # Expect declarations
│       │   │   └── presentation/    # Screens, components, themes
│       │   └── sqldelight/          # Database schema
│       ├── androidMain/             # Android implementations
│       │   └── kotlin/
│       │       ├── MainActivity.kt
│       │       ├── VisionLearnApplication.kt
│       │       └── platform/        # TTS, ImagePicker, etc.
│       └── iosMain/                 # iOS implementations
│           └── kotlin/
│               ├── MainViewController.kt
│               └── platform/        # TTS, ImagePicker, etc.
├── iosApp/                          # Xcode project
│   ├── iosApp/
│   │   ├── iOSApp.swift
│   │   ├── ContentView.swift
│   │   └── Info.plist
│   └── iosApp.xcodeproj/
├── gradle/
│   └── libs.versions.toml           # Version catalog
├── screenshots/                     # App screenshots
├── build.gradle.kts
├── settings.gradle.kts
└── local.properties                 # API keys (gitignored)

The Perkins CVI Protocol

This app is designed around the Perkins CVI Protocol, a comprehensive educational assessment tool created by The CVI Center at Perkins School for the Blind. The protocol identifies 16 Visual Behaviors:

Visual Attention
Visual Recognition
Visual Curiosity
Object Complexity
Array Complexity
Sensory Complexity
Light Preference
Color Preference
Movement Need
Visual Latency
Visual Field Preference
Distance Viewing
Visual-Motor Integration
Visual Reflexes
Novelty Tolerance
Face Recognition

VisionLearn focuses on Color Preference and Object/Array Complexity - key factors that can be addressed through software adaptation.

Learn more: Perkins CVI Protocol

Future Steps

Additional Visual Behaviors - Support more of the 16 behaviors
AI Image Simplification - Automatically simplify complex images
Switch Access - Full switch device support for motor impairments
Multi-language Support - Localization for global accessibility
Cloud Sync - Sync profiles and progress across devices
Therapist Dashboard - Web portal for professionals
Export Reports - Generate progress reports for IEP meetings

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Acknowledgments

Perkins School for the Blind for the CVI Protocol framework
JetBrains for Kotlin Multiplatform
Google for Gemini AI

Created for the KotlinConf 2026 Kotlin Multiplatform Contest

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
composeApp		composeApp
gradle		gradle
iosApp		iosApp
screenshots		screenshots
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
build.gradle.kts		build.gradle.kts
gradle.properties		gradle.properties
gradlew		gradlew
settings.gradle.kts		settings.gradle.kts

License

arulagarwal/VisionLearn

Folders and files

Latest commit

History

Repository files navigation

VisionLearn

Table of Contents

Motivation and Background

What is CVI?

The Problem

Features

Visual Profile System

Five Learning Modules

Accessibility First

AI-Powered Learning (Google Gemini)

Custom Activity Creator

Progress Tracking

Screenshots

Onboarding Flow

Home & AI Generation

Learning Modules

Progress & Tools

iOS Version

Demo Video

Technical Implementation

Architecture

Shared Code Structure

Platform-Specific Code

AI Integration

How to Run

Prerequisites

1. Clone the Repository

2. Configure Gemini API Key (Optional but Recommended)

3. Build & Run

Android

iOS

Troubleshooting

Project Structure

The Perkins CVI Protocol

Future Steps

License

Acknowledgments

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages