VIOVIO Survey BenchmarkGitHub Pages reproducible benchmark

Lesson 01 / Beginner

What VIO estimates

Learn what visual-inertial odometry produces and what it does not claim to solve.

Learning outcomes

  • Explain pose, velocity, bias, and trajectory output
  • Separate odometry output from map-building claims
  • Read a pose row without guessing units or frame convention

Beginner explanation

VIO estimates how a camera-IMU platform moves over time. The benchmark output is a sequence of poses, one per timestamp.

A pose combines position and orientation. The public output format is timestamp tx ty tz qx qy qz qw.

The benchmark compares odometry trajectories. It does not reward future-corrected or globally optimized outputs as primary results.

Expert checks

Global translation and yaw can be gauge freedoms depending on sensing and alignment. EPA metadata must state the alignment mode.

Velocity and IMU bias are often internal state variables even when only pose is exported for evaluation.

Practice task

Open any result row and identify the expected trajectory path, timestamp unit, pose convention, and EPA output files.