Railways play a pivotal role in green transportation due to their energy efficiency and lower carbon footprint compared to road and air transport. Integrating Artificial Intelligence (AI) into railways is revolutionizing this sector, especially in autonomous driving. While AI integration in railways significantly contributes to the green transportation initiative, it faces notable challenges due to diverse weather and illumination conditions. These varying environmental factors can impede AI models' ability to accurately and consistently recognize and react to different scenarios, hindering the reliability of autonomous train operations. This limitation in AI's adaptability and generalization across different "theatre conditions" underscores the need for more robust and adaptable AI solutions to ensure the effective and safe implementation of autonomous railway systems.
Goal: The aim of this contest is to support the detection of different weather events that could lead to an inaccurate object detection in autonomous railway.
The dataset consists of .PNG samples representing six different weather and scene illumination conditions in the context of smart railways. In particular, all the pictures have been generated by using MathWorks' RoadRunner, simulating an RGB camera mounted in front of the train, with no obstacles on rail tracks.
The class labels are as follows:
CL - Stands for Clean and is associated with label 0
BR - Stands for Bright Right and is associated with label 1
DA - Stands for Darkness and is associated with label 2
RA - Stands for Rainy and is associated with label 3
SF - Stands for Sun Flare (Direct Sunlight Interference) and is associated with label 4
SH - Stands for Shadows and is associated with label 5
The dataset can be downloaded from this link: https://www.kaggle.com/datasets/ifruit94/weatherdataset