Vanderwaals

<div align="center"> <img style="display: block; border-radius: 9999px;" src="Vanderwaals_logo_black.png" width="500" alt="Vanderwaals Logo"> <h1>Vanderwaals</h1> <p> <img src="https://img.shields.io/github/v/release/avinaxhroy/Vanderwaals?style=for-the-badge" alt="Release"> <img src="https://img.shields.io/github/downloads/avinaxhroy/Vanderwaals/total?style=for-the-badge" alt="Downloads"> <img src="https://img.shields.io/github/license/avinaxhroy/Vanderwaals?style=for-the-badge" alt="License"> </p> <p><strong>AI-Powered Wallpaper App That Learns Your Aesthetic</strong></p> <p><i>Next-generation wallpaper personalization powered by MobileNetV4</i></p> </div>

---

✨ What is Vanderwaals?

Vanderwaals is a cutting-edge Android wallpaper application that leverages state-of-the-art on-device machine learning to understand your unique aesthetic preferences and automatically curate wallpapers you'll love. Powered by MobileNetV4-Conv-Small, it delivers unparalleled personalization without compromising your privacy.

🎯 Key Highlights

• 🤖 Advanced ML Engine: MobileNetV4-Conv-Small neural network with 1280-dimensional embeddings for superior aesthetic understanding
• 📚 Massive Curated Library: 6,000+ wallpapers from GitHub collections + 5,400+ from Bing's photography archive
• 🔒 Privacy-First: 100% offline ML processing, zero analytics, no data collection
• ⚡ Lightning Fast: 9x faster wallpaper changes (~5s) with intelligent pre-caching and chunked processing
• 🎨 Revolutionary Glassmorphism: True Apple-style liquid glass effect that Jetpack Compose can't natively achieve—engineered around platform limitations
• 📦 90% Smaller Downloads: Quantized embeddings reduce manifest size from 60MB to ~6MB

---

🆕 What's New in v4.5.0

MobileNetV4 Upgrade 🚀

• 1280D Embeddings: Upgraded from MobileNetV3 (576D) to MobileNetV4-Conv-Small for 2.2x more detailed aesthetic understanding
• 80% Faster Inference: Optimized architecture delivers faster on-device processing
• Improved Accuracy: Better recognition of artistic styles, moods, and compositional elements

Performance Improvements ⚡

• 9x Faster Changes: Wallpaper change time reduced from ~45s to ~5s through chunked processing
• Memory Efficient: Peak memory usage reduced from 190MB to ~40MB with batched operations
• Pre-Caching: Second wallpaper change is near-instant with background computation
• 90% Smaller Downloads: Quantized embeddings reduce manifest from 60MB to ~6MB

Enhanced Features ✨

• Fast Initialization: First like immediately shapes recommendations (no more waiting)
• Instant Dislike: Automatically applies new wallpaper when you dislike one
• Flexible Intervals: New 3h, 6h, 12h auto-change options
• Daily Playlist: 15 pre-selected wallpapers for "Every Unlock" mode
• Smart Migration: Seamless upgrade from v3.x with user-friendly dialog

Reliability & Stability 🛡️

• Samsung Optimized: WakeLock and foreground service for One UI battery restrictions
• Auto-Change Fix: Works reliably even after app swipe (foreground service)
• Android 15+ Boot Fix: Resolved ForegroundServiceStartNotAllowedException crashes
• Smart Crop 2.0: Lossless PNG caching and resolution preservation

UI Polish 🎨

• True Liquid Glass: Revolutionary glassmorphism technique that solves Compose's blur limitations
  • Pre-rendered backgrounds with chromatic aberration and barrel distortion
  • Slice-based rendering for pixel-perfect glass cards
  • 80px Gaussian blur + RGB separation for authentic Apple-style frosted glass
• Enhanced Glassmorphism: Multi-layer blur effects and glassmorphic cards throughout
• Light Mode: Full light theme support for all screens
• Bing Selection UI: Interactive "Radio Cards" for meaningful source selection
• Enhanced Analytics: Personalization insights and learning progress tracking

---

Screenshots

<div align="center"> <br/> <img style="display: block; border-radius: 9999px;" src="Screenshots/1.png" width="150" alt="Free"> <img style="display: block; border-radius: 9999px;" src="Screenshots/2.png" width="150" alt="Beautiful Wallpapers"> <img style="display: block; border-radius: 9999px;" src="Screenshots/3.png" width="150" alt="Upload Favourite"> <img style="display: block; border-radius: 9999px;" src="Screenshots/4.png" width="150" alt="Auto-Change Wallpaper"> <br/> <img style="display: block; border-radius: 9999px;" src="Screenshots/5.png" width="150" alt="Personalization Insight"> <img style="display: block; border-radius: 9999px;" src="Screenshots/6.png" width="150" alt="Tap and Learn"> <img style="display: block; border-radius: 9999px;" src="Screenshots/7.png" width="150" alt="More Algorithm"> <br/> </div>

📥 Download

<div align="center"> <a href="https://github.com/avinaxhroy/Vanderwaals/releases"> <img src="https://raw.githubusercontent.com/Kunzisoft/Github-badge/main/get-it-on-github.png" alt="Download from GitHub" height="80"> </a> <a href="https://play.google.com/store/apps/details?id=me.avinas.vanderwaals"> <img src="https://camo.githubusercontent.com/2ea3b1d605be783034a4a4cf364f24cdd29b52f60caf0892b03c3e53e85b91ce/68747470733a2f2f706c61792e676f6f676c652e636f6d2f696e746c2f656e5f75732f6261646765732f7374617469632f696d616765732f6261646765732f656e5f62616467655f7765625f67656e657269632e706e67" alt="Download from GitHub" height="80"> </a> </div>

Requirements

• Android 12.0 (API 31) or higher
• ~50MB storage space (app + cache)
• Internet connection for initial wallpaper sync (then works offline)

---

🚀 Features

🎨 Intelligent Personalization

Auto Mode

Start fresh with algorithm-selected wallpapers from curated collections. The app learns your taste as you provide feedback with lightning-fast initialization—your first like immediately shapes recommendations.

Personalize Mode

Upload one favorite wallpaper and instantly get 100+ similar matches. The advanced ML algorithm analyzes:

• Deep Visual Features (70%): MobileNetV4 1280D embeddings capture artistic style, composition, and mood with 2.2x more detail than previous models
• Color Palette (20%): LAB color space analysis for perceptually accurate matching
• Category Affinity (10%): Learns your preference for categories (gruvbox, nord, nature, minimal, etc.)

🧠 Advanced Learning Algorithm

• MobileNetV4-Conv-Small: State-of-the-art 1280-dimensional embeddings for superior aesthetic understanding
• Fast Initialization: First like/download immediately sets preference baseline for instant personalization
• Exponential Moving Average (EMA): Smoothly integrates new preferences without forgetting old ones
• Implicit Feedback: Learns from wallpaper duration (quick changes = dislike, long duration = like)
• Temporal Diversity: Prevents repetitive categories with intelligent recency penalties
• Progressive Trust: Balances original embeddings with learned preferences over time

📚 Rich Content Sources

• GitHub Collections (6,000+ wallpapers):

  • dharmx/walls - Curated aesthetic wallpapers
  • D3Ext/aesthetic-wallpapers - Minimalist designs
  • makccr/wallpapers - Linux ricing wallpapers
  • Gingeh/wallpapers - Anime aesthetics
  • FrenzyExists/wallpapers - Dark themes
  • Plus additional curated collections

• Bing Photography (5,400+ archive):

  • Bing Lite: ~1,000 wallpapers from last 3 years
  • Bing Full: Complete archive from 2009-present
  • Daily featured wallpapers in UHD quality
  • Professional photography from around the world
  • Full MobileNetV4 embedding support

• Smart Auto-Sync: Automatic weekly content updates from all sources with incremental updates

• 90% Smaller Downloads: Quantized embeddings reduce manifest from 60MB to ~6MB

⚙️ Powerful Automation

• Auto-Change Modes:

  • Every device unlock (with Daily Playlist system—15 pre-selected wallpapers for instant changes)
  • Fixed intervals (1h, 3h, 6h, 12h, 24h)
  • Daily at custom time
  • Manual only

• Apply To:

  • Lock screen
  • Home screen
  • Both screens simultaneously

• Smart Features:

  • Pre-Caching: Background computation for near-instant second changes
  • Instant Dislike: Automatically applies new wallpaper when you dislike one
  • Samsung Optimized: Dedicated "Keep-Alive" service with WakeLock for One UI battery restrictions
  • Reliable Background Processing: Foreground service ensures consistent auto-change even after app swipe
  • Smart Crop: Intelligent image cropping using saliency detection to focus on interesting regions

📊 History & Analytics

• Chronological Timeline: View all applied wallpapers with timestamps
• Quick Actions: Like ❤️, Dislike 👎, or Download wallpapers directly from history
• Smart Grouping: Date-based sections (Today / Yesterday / Month Year)
• Full-Screen Preview: Zoomable image viewer for detailed inspection
• Personalization Insights: Track learning progress and category preferences

🎨 Modern UI/UX

• True Liquid Glass Effect: Revolutionary Apple-style glassmorphism that Jetpack Compose can't natively achieve
  • Pre-rendered backgrounds with chromatic aberration and edge distortion
  • Slice-based rendering - cards display perfectly aligned background slices
  • Multi-layer processing: Heavy Gaussian blur (80px) + chromatic RGB separation + barrel distortion
  • Smart Launcher technique: Process once, render infinitely - solves Compose's real-time blur limitations
  • Opaque illusion: Cards are solid