VIOVIO Survey BenchmarkGitHub Pages reproducible benchmark

Lesson 03 / Practitioner

Run one sequence reproducibly

Move from one dataset sequence to a standard benchmark output folder.

Learning outcomes

  • Choose a compatible final public system
  • Run with a recorded Git commit and configuration
  • Save pose, timing, resource, and command metadata

Beginner explanation

Start with one known sequence before trying a batch run. A single successful case validates the data path and output format.

Do not edit result tables by hand before the trajectory and EPA outputs exist.

Expert checks

Record current-timestep odometry output. If a system has SLAM-specific correction modules, the benchmark configuration must use odometry-only reporting.

Store exact command, container tag, host platform, dataset conversion, and config file with the output artifacts.

Practice task

Create results/pose/<system>/<dataset>/<sequence>/poses.txt and confirm every row has eight fields.