NVIDIA HDR Converter GUI

Convert NVIDIA HDR screenshots (JXR format) to JPEG with advanced AI enhancement and intelligent tone mapping.

Features

Core Functionality

• Convert NVIDIA JXR (HDR) screenshots to JPEG format
• Intelligent HDR tone mapping with automatic algorithm selection
• AI-powered color enhancement using ensemble deep learning models
• Batch processing support for folders
• Live preview with before/after comparison and histogram visualization
• HDR metadata extraction from JXR files for optimal processing

AI Color Enhancement

• Multi-model ensemble using VGG16, ResNet34, and DenseNet121
• CBAM attention mechanism for spatial and channel-wise feature enhancement
• Perceptual color preservation during tone mapping
• Edge enhancement with Sobel filters and adaptive strength control
• Color balance optimization across shadows, midtones, and highlights
• Half-precision (FP16) support for memory-efficient GPU processing

Advanced Tone Mapping Algorithms

• Perceptual - Advanced tone mapping preserving local contrast and color relationships
• Adaptive - Intelligent blending of multiple operators based on image regions
• Mantiuk06 - Contrast-based tone mapping with local adaptation
• Drago03 - Logarithmic tone mapping optimized for extreme dynamic ranges
• Hable - Filmic tone mapping curve (Uncharted 2)
• ACES - Academy Color Encoding System RRT+ODT
• Reinhard - Extended Reinhard with white point adaptation
• Filmic - Cinematic tone mapping
• Uncharted2 - Game-optimized tone mapping

Intelligent Processing

• Automatic tone mapping selection based on comprehensive image analysis
• Scene classification (high-key, low-key, extreme highlights detection)
• Dynamic range analysis with histogram-based optimization
• Local contrast preservation and enhancement
• Color saturation analysis and adaptive correction

User Interface

• Modern dark-themed GUI with TKinterModernThemes
• Dual preview modes (Big/Small) with resizable interface
• Real-time histogram visualization with RGB channel analysis
• Progress tracking for batch operations
• Device switching (GPU/CPU) with live performance monitoring
• Parameter adjustment with live preview updates

Requirements

System Requirements

• Python 3.8+
• NVIDIA GPU (recommended) with CUDA support
• 4GB+ RAM (8GB+ recommended for large images)

Python Dependencies

torch>=2.0.0
torchvision>=0.15.0
Pillow==10.2.0
numpy==1.26.4
matplotlib==3.9.3
imagecodecs==2024.9.22
TKinterModernThemes==1.10.4

Install PyTorch: Visit https://pytorch.org/ for CUDA-compatible installation

Installation

1. Clone the repository:

   git clone https://github.com/5ymph0en1x/NVIDIA-HDR-Converter-GUI.git
   cd NVIDIA-HDR-Converter-GUI

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   python NHC.py

Usage

Single File Conversion

1. Launch the application
2. Select "Single File" mode
3. Browse and select input JXR file
4. Choose output JPEG location
5. Adjust parameters (gamma, exposure, enhancement strength)
6. Click "Convert"

Batch Processing

1. Select "Folder" mode
2. Choose folder containing JXR files
3. Configure enhancement parameters
4. Click "Convert"

Note: Converted files are saved in a Converted_JPGs subfolder with automatic naming

Parameter Configuration

Basic Parameters

• Tone Map: Displays auto-detected optimal algorithm
• Gamma: Pre-gamma correction (0.1-3.0, default: 1.0)
• Exposure: Auto-exposure adjustment (0.1-3.0, default: 1.0)

AI Enhancement Controls

• Enable AI Enhancement: Toggle neural network-based color correction
• Strength: Edge enhancement intensity (0-100%, default: 50%)

Device Settings

• GPU/CPU Selection: Automatic CUDA detection with fallback
• Half-Precision Toggle: Enable FP16 for memory efficiency (GPU only)

Interface Modes

• Big Mode: 2200×850 window with 720×406 previews
• Small Mode: 1760×840 window with 512×288 previews

Technical Architecture

HDR Processing Pipeline

1. JXR Decoding with metadata extraction
2. Image Analysis for optimal tone mapping selection
3. Tone Mapping using selected algorithm
4. AI Enhancement (optional) with color correction
5. Edge Enhancement (optional) with adaptive filtering
6. sRGB Conversion and JPEG encoding

AI Enhancement Architecture

Input → VGG16 Features    ↘
     → ResNet34 Features  → Feature Fusion → CBAM Attention → Color Transform → Output
     → DenseNet Features  ↗

Key Components

• HDRMetadata: Extracts luminance and color space information
• AdvancedToneMapper: Multi-algorithm tone mapping with automatic selection
• PerceptualColorPreserver: CIE LAB color space preservation
• ColorCorrectionNet: Ensemble deep learning model
• EdgeEnhancementBlock: Sobel-based edge detection and enhancement
• DeviceManager: GPU/CPU switching with memory optimization

Performance Optimizations

• Memory-efficient tensor operations with automatic cleanup
• Multi-threaded processing for batch operations
• Progressive image loading for large files
• CUDA memory management with automatic cache clearing
• Half-precision support reducing memory usage by 50%

Image Analysis Metrics

• Dynamic range analysis (min/max/mean luminance)
• Zone system analysis (shadow/midtone/highlight distribution)
• Local contrast measurement with variance analysis
• Color saturation evaluation across channels
• Scene classification (high-key/low-key/extreme highlights)

Supported Formats

Input

• JXR files (JPEG XR) from NVIDIA HDR screenshots
• HDR metadata automatic extraction and utilization

Output

• JPEG with optimized quality settings (95% quality, optimized compression)

Logging and Debugging

• Comprehensive logging to hdr_converter.log
• Real-time status updates in GUI
• Error handling with detailed messages
• Performance metrics logging

Known Limitations

• JXR format only - other HDR formats not supported
• NVIDIA screenshots - optimized for NVIDIA HDR capture format
• Memory requirements - large images may require significant RAM

Troubleshooting

Common Issues

• "imagecodecs version compatibility": Update imagecodecs to latest version
• CUDA out of memory: Enable FP16 mode or switch to CPU processing
• JXR decode failure: Ensure file is valid NVIDIA HDR screenshot

Performance Tips

• Use GPU mode for faster processing
• Enable FP16 to reduce memory usage
• Close other applications when processing large batches
• Use Small mode for lower memory usage

Acknowledgments

• PyTorch pretrained models (VGG16, ResNet34, DenseNet121)
• TKinterModernThemes for modern GUI design
• imagecodecs for JXR decoding support
• NVIDIA for HDR screenshot format documentation

License

This project is open source. Please check the license file for details.

---

Version: Enhanced Edition with AI-powered processing and intelligent tone mapping