SafeSpace 🛡️👶

SafeSpace is an intuitive, interactive iOS application designed to help parents and guardians identify and mitigate potential household dangers for children in real-time. Built specifically for the Apple Swift Student Challenge, SafeSpace leverages the power of on-device Machine Learning and Augmented Reality to turn an iPhone into a smart safety scanner.

🌟 Features

• Real-Time Hazard Detection: Simply point your camera around your home. SafeSpace's custom CoreML model instantly processes the live feed to identify hazards.
• Dynamic AR Overlays: When a dangerous item is detected, SafeSpace overlays a smooth, pulsing AR bounding box directly onto the physical object, making the danger immediately obvious.
• Smart Scanning System: Avoids jarring false positives with a built-in 1.2-second confident scanning ring before alerting the user.
• Actionable Safety Tips: Displays clear, non-alarming, and actionable safety advice on a neat "Danger Card" corresponding to the specific hazard detected.
• History Tracking: Automatically logs the last 20 detected hazards along with their timestamps and confidence levels so you can review potential blind spots in your home later.
• Privacy First, On-Device Processing: All computer vision and machine learning inferences happen locally on the device via the Vision framework—no camera feeds are ever sent to the cloud.

🎯 What We Detect

SafeSpace's model is trained to recognize three primary high-risk categories in the home:

1. ✂️ Scissors & Knives (Sharp objects)
2. 🔌 Electrical Outlets & Cables (Exposed sockets and wiring)
3. 🫧 Cleaning Products (Harmful chemicals within reach)

🛠️ Technology Stack

SafeSpace is built using modern Apple frameworks and fully embraces declarative UI and spatial computing concepts:

• SwiftUI: Entirely built with SwiftUI for a reactive, state-driven, and highly animated interface (including custom bottom sheets, onboarding flows, and haptic feedback integrations).
• CoreML & CreateML: Uses a custom trained Image Classification model natively integrated to run optimally on Apple Silicon.
• ARKit & SceneKit: Uses ARWorldTrackingConfiguration and ARSCNView wrapped in a UIViewRepresentable to project safety warnings spatially.
• Vision Framework: Handles the heavy lifting of extracting and processing CVPixelBuffer frames from the ARKit camera feed efficiently on background queues.
• Combine: Powers the reactive ARViewModel pipeline to keep the view state perfectly synchronized with hardware sensors.

🚀 Getting Started

Prerequisites

• Xcode 15.0 or later
• iOS 17.0 or later
• An iPhone or iPad with an A9 chip or later (Requires physical device for ARKit camera tracking; the iOS Simulator does not support live ARKit camera feeds).

Installation

1. Clone the repository:

   git clone https://github.com/tade-dev/FoodSenseAR.git

2. Open FoodSenseAR.xcodeproj in Xcode.
3. Select your physical iOS device as the run destination.
4. Ensure your Apple Developer Signing Identity is selected in the Signing & Capabilities tab.
5. Build and Run (Cmd + R).

💡 How It Works

1. Onboarding: First-time users are greeted with a beautiful, paginated onboarding flow explaining house scanning.
2. Scanning: As you pan the camera, the ARViewContainer passes frames every 0.5s to the DetectionService.
3. Validating: If the CoreML model hits a confidence threshold of >90%, a visual HUD ring begins tracking the object.
4. Alerting: If the object remains in frame for an uninterrupted 1.2 seconds, the AR engine drops a frosted-glass Danger Card providing detailed safety instructions, and logs it to History.

---

Created with ❤️ for the Apple Swift Student Challenge.