Images
A Glance over the EuroCity Person Dataset
4 seasons, 12 countries, 31 cities, 47.300 images, 238.200 persons. Accurate annotations with object bounding boxes.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Beyond Bounding Boxes
Apart from bounding boxes, body orientations of pedestrians and riders were annotated, as well as additional tags such as different occlusion and trunction levels.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
Videos
A Glance over the EuroCity Person Dataset
4 seasons, 12 countries, 31 cities, 47.300 images, 238.200 persons. Accurate annotations with object bounding boxes.
Data Recordings
Images were recorded every 4 second and annotated. Here a fragment from the Eternal City of Rome (blurring due to privacy concerns on publicly accessible part of website).
Adverse Light and Weather Conditions
Apart from dry weather at daytime, the dataset covers low frontal sun, rain, snow and night-time conditions.
Crowds and Heavy Occlusions
The dataset also features challenging scenes with crowds and heavy occlusions
Beyond Bounding Boxes
Apart from bounding boxes, body orientations of pedestrians and riders were annotated, as well as additional tags such as different occlusion and trunction levels.
Detailed Pose Annotations (ECPDP Extension)
In addition to bounding box annotations and class information as in the ECP detection dataset, the ECPDP package provides detailed pose annotations consisting of 17 joint points. For every joint, it is indicated if it is fully visible, self-occluded or occluded.