Abstract

Understanding traffic scene images taken from vehicle mounted cameras provides important information for high level tasks such as autonomous driving and advanced driver assistance. The problem is far from trivial especially due to challenges from weather and illumination variation. To facilitate the research against such challenges, we present a new benchmark for cross-weather traffic scene understanding. The dataset consists of 1,356 traffic scene images collected at 226 different locations. For each location, there are six images taken by a vehicle mounted camera under different weather/illumination conditions including sunny day, night, snowy day, rainy night, cloudy day I and cloudy day II. We manually annotated each image with scene understanding labels such as road, sky, building, etc. To the best of our knowledge, this is the first carefully collected benchmark for cross-weather traffic scenes.

Reference

A Benchmark for Cross-Weather Traffic Scene Understanding Shuai Di, Honggang Zhang, Xue Mei, Danil Prokhorov, and Haibin Ling International Conference on Intelligent Transportation Systems (ITSC), 2016.

Dataset

How to download the benchmark?

All the data including original images and human annotation can be downloaded entirely Cross-Weather.zip.

How to use the data?

Txt file listing classes and label colors as RGB triples.

The data in Cross-Weather.zip contain the following folders:

• six folders include the original images of each scenario, respectively.
• another six folders include the human annotation corresponding to the original images of each scenario.

 

 

The Foggy-Sunny data.