Skip to content
 
 

Latest commit

 

History

222 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🌌 ZeroVision — High-Performance Image Processing & AI Studio

Type: Desktop Application Ecosystem Platform Distribution: Standalone Single-File Build and Test License: Apache 2.0

ZeroVision (incorporating Aurora Studio) is a desktop workstation application (WPF, .NET 8) for high-fidelity image management, color grading, and computer vision enhancement. Part of the sovereign ZeroUniverse ecosystem, it combines a non-destructive 32-bit float Linear Light develop pipeline (similar to Adobe Lightroom / Darktable) with autonomous DirectML ONNX AI engines for super-resolution upscaling, face restoration, and intelligent auto-tagging.


ZeroVision Preview


🌟 Key Capabilities

1. Non-Destructive Develop Pipeline (Linear Light)

All adjustments operate in linear light float RGBA color space through an intelligent DAG pipeline, preventing color banding and preserving maximum highlight/shadow dynamic range.

Tone & Lighting

  • Exposure, Contrast, Highlights, Shadows, Whites, Blacks
  • Tone Curve & Parametric Curve with interactive drag-point editing and presets (Linear, Medium, Strong, Faded)
  • Filmic, Filmic RGB, Sigmoid, Tone Equalizer, Dehaze
  • Levels (per-channel histogram stretching, Auto Levels, Auto Color for color cast removal)
  • Interactive Histogram with direct tone adjustment by dragging

Color & Grading

  • White Balance (Kelvin temperature slider, Auto WB, eyedropper tool, standard illuminant presets)
  • HSL 8-Channel Mixer + Targeted Adjustment Tool (TAT) (click-drag directly on image canvas for rapid HSL tuning)
  • Color Balance RGB 4-Way and Color Grading wheels
  • Split Toning, Channel Mixer, Selective Color, Color Unify, Velvia, Color Contrast (Lab)
  • 3D LUT (.cube) support and ICC input color profiles (sRGB, AdobeRGB, Rec2020, Display P3)
  • Black & White: Deep channel mixing with classic color filters and toning simulation
  • Film Negative (negadoctor): High-precision negative film scan inversion and color reconstruction

Detail & Sharpness

  • Sharpen (radius control + intelligent edge Masking)
  • Noise Reduction (Luminance, Color, Chroma)
  • Diffuse-or-sharpen (PDE) filter, Hot Pixel removal, CA Correct, Defringe
  • Texture, Clarity, Grain (monochrome and chromatic film grain synthesis)

Geometry & Layout

  • Crop with free or standard aspect ratios (1:1, 16:9, 4:3, golden ratio) and composition guide overlays
  • Straighten, Rotate, Flip with EXIF-based auto-rotation
  • Perspective / Upright correction, Liquify/Warp with intuitive handle-based editing
  • Lens Correction: Automatic distortion and vignetting correction via lensfun database (EXIF-based) or manual adjustment

Local Adjustments & Masking

  • Linear Gradient Mask, Radial Mask, Brush Mask
  • Luminance Mask and Color Range Mask for targeted tonal edits
  • Mask inversion, feathering, opacity, and multi-mask stacking

2. Autonomous AI & Vision Pipeline

ZeroVision decouples heavy AI inference into isolated background host processes (ImageTool.Host) via DirectML / ONNX Runtime:

AI Plugin Model Architecture Role
Upscaler Real-ESRGAN / AuraSR 2x / 4x super-resolution upscaling with DirectML acceleration
Face Restorer GFPGAN / CodeFormer High-fidelity portrait reconstruction and eye/facial feature enhancement
Vision Tagger MobileNet / ResNet ONNX Semantic multi-label scene and object classification

🏗 Architecture & Project Layout

ZeroVision/
├── ImageTool.Core/          # Domain models, preset system, pipeline abstractions
├── ImageTool.Imaging/       # 32-bit float linear light image processing kernels
├── ImageTool.Host/          # Out-of-process DirectML ONNX AI runner
├── ImageTool.Shared/        # IPC contracts, cross-process shared memory buffers
├── ImageTool.Plugins.*/     # Autonomous plugin modules (Upscaler, FaceRestorer, VisionTagger)
├── TestDML/                 # DirectML hardware acceleration diagnostic harness
└── ImageTool.Tests/         # ~800 automated unit and integration tests

⚡ Quick Start for Developers

Prerequisites

  • Windows 10 (build 1809+) or Windows 11 (64-bit)
  • .NET 8 SDK
  • DirectX 12 capable GPU (for DirectML AI acceleration)

Build and Run

# Clone repository
git clone https://github.com/kzxl/ZeroVision.git
cd ZeroVision

# Build solution
dotnet build ImageTool.slnx -c Release

# Run automated test suite
dotnet test ImageTool.Tests/ImageTool.Tests.csproj

Packaging

To generate optimized Lite and Full deployment packages:

pwsh ./publish.ps1

📄 License

Licensed under the Apache License 2.0. Part of the sovereign ZeroUniverse industrial computing ecosystem.

About

A high-performance WPF .NET 8 application for AI Image Upscaling (4x-UltraSharp) and Face Restoration (GFPGAN), featuring DirectML hardware acceleration.

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages