VIOVIO Survey BenchmarkGitHub Pages reproducible benchmark

Lesson 02 / Beginner + Practitioner

Sensors, timestamps, and calibration

Understand the input package before running any method.

Learning outcomes

  • Identify image, IMU, calibration, and ground-truth files
  • Check timestamp units and monotonic order
  • Recognize camera-IMU extrinsic and temporal assumptions

Beginner explanation

A VIO dataset needs images, IMU measurements, timestamps, and calibration. Missing or inconsistent calibration can dominate the result.

The IMU is high-rate. It bridges rapid motion between camera frames and helps estimate orientation and scale.

Ground truth is used by EPA; it is not an input to the VIO system during a normal benchmark run.

Expert checks

Record camera model, distortion model, camera-IMU transform, time offset handling, IMU noise, gravity convention, and body frame.

Do not compare resource or accuracy rows if dataset conversion changes sensor timing without recording it.

Practice task

Before running a sequence, write down image rate, IMU rate, timestamp unit, camera model, and camera-IMU transform source.