Galway Multimodal Infrastructure Node Dataset

Autonomous vehicles perceive the world through onboard sensors that, when occluded by infrastructure or other vehicles, provide an incomplete dynamic environmental map (DEM), leading to lower safety and occupant comfort. In an ideal intelligent transportation system (ITS), infrastructure nodes create an accurate and reliable DEM and publish this to the surrounding vehicles with infrastructure-to-vehicle (I2V) communication. While I2V communication links are a heavily researched topic, research on creating a DEM from an infrastructure node has somewhat stagnated due to a lack of adequate datasets. In this paper, we present a new dataset, G-MIND, that allows researchers to determine the sensor or suite of sensors that deliver the accuracy and reliability required to support safety-critical ITS applications such as cooperative collision avoidance. The proposed dataset consists of data from 10 different sensors, including; cameras, LIDARs, RADAR, and an event-based camera. The dataset has 86,400 annotated frames containing cars, pedestrians, and bicycles.