NLoS Benchmark

Welcome to the Non-Line-of-Sight Imaging Benchmark!

On this website, we provide reference data and evaluation tools that aim to make existing NLoS reconstruction algorithms more comparable. The website is accompanied by a paper published at the 2018 British Machine Vision Conference.

You are free to use the data and tools provided here to benchmark your own non-line-of-sight reconstruction algorithms. The benchmark is published by the Computational Light Transport / Digital Material Appearance Group within the Institute of Computer Science at the University of Bonn.

In 2025, the NLoS challenge was discontinued. All participation features (including registration, login, submission of results, evaluation, and ranking) have been disabled and all user data deleted.

We keep this website online for the forseeable future in the hope that the dataset and transient image tools will be helpful for other researchers.

 

Presentation at the BMVC

04.09.2018

Our benchmark is published at the BMVC 2018 today. Check out our poster during the poster session and have a talk with the authors!

The challenges

Geometry Reconstruction

What is the shape of the hidden object?

Object Tracking

How does the object move, and what is its orientation?

Object Classification

From a set of known objects, which one is currently hidden in the scene?

Texture Reconstruction

How is the hidden object painted?