ntu-aris / MMAUD

This package provides A Comprehensive Multi-Modal Anti-UAV Dataset
MIT License
38 stars 7 forks source link

Why can't the number of images align with GT and the file names don't seem to correspond one-to-one either #1

Closed lyq-y closed 2 weeks ago

lyq-y commented 8 months ago

Hello dear author, thank you very much for your open source work, I have encountered some problems working with your dataset, including: First of all, if it is based on a2D object detection task, what does the3D vector in GT mean? On the other hand, the number of GTs and file names don't seem to correspond one-to-one with images. best wishes to you!

snakehaihai commented 6 months ago

Hi

The 3D vector in the GT denotes the tracked pose of the UAV with respect to the Leica Frame. Leica is one of the most accurate tools for outdoor precision trajectory estimation. This measurement uncertainty should be less than 1 cm.

It is common as each camera is running 30hz. LIDARs are 10hz and ground truth is only 10hz. Usually you don't need to worry about the stamp differences as the result trajectory you predicted can be used to compare with the ground truth trajectory using EVO https://github.com/MichaelGrupp/evo

I`ll upload an other set of evaluation pipeline script in next month.

image

Rgs

Shenghai Yuan