paper-benchmark-execution

What problem does it solve?

This plan coordinates and executes the paper benchmark workflow for RGB, multichannel, DataLoader, and video benchmarks, enabling reproducible experiments and consistent result recording.

Core Features & Use Cases

• Orchestrates the paper benchmark execution using the canonical sources at _internal/plans/paper_benchmark_execution_plan.md and the policy and matrix definitions in benchmark.
• Validates and applies checked-in YAML configs under configs/paper to drive CPU and GPU runs across RGB, 9-channel, and video scenarios.
• Supports re-running, result validation, and comparison across libraries (e.g., AlbumentationsX, torchvision, Kornia) within the prescribed plan.

Quick Start

Run the paper benchmark execution workflow by selecting the appropriate config from configs/paper and invoking the plan and run commands as documented.