Benchmark on Inference for Large-Scale Inverse Problems
Getting Started
Quickstart
Environment Setup
Installation
Running the Benchmark
Viewing Results
Configuration Guide
Available Configuration Files
Parallel Configuration (SLURM)
Main Configuration (Experiments)
Customizing Your Configuration
Examples
High-Resolution Color Image Deblurring
The Dataset: High-Resolution Color Image
Configuration: Experiment Setup
Execution Grid
Dataset Parameters
Solver
Interpreting Results
Interpretation
2D Tomography Reconstruction
The Dataset: Shepp-Logan Phantom
Configuration: Experiment Setup
Execution Grid
Dataset Parameters
Solver
Interpreting Results
Interpretation
3D Cone-Beam CT Reconstruction
The Dataset: Walnut Cone-Beam CT
Configuration: Experiment Setup
Execution Grid
Dataset Parameters
Solver
Interpreting Results
Interpretation
Key Takeaways
GPU Performance Under the Hood
Image Shape Effects on GPU Memory Performance
Batch Size Effects on Runtime
Distributed Scaling Insights
Overview
High-Resolution Color Image
Tomography 2D
Tomography 3D
Benchmark on Inference for Large-Scale Inverse Problems
Key Takeaways
View page source
Key Takeaways
Insights and best practices derived from the benchmark results.
GPU Performance Under the Hood
Distributed Scaling Insights