The Egocentric Perspective and Overhead-View Segmentation from Honda (EPOSH) dataset consists of 481 densely annotated ego-centric road-scene video clips collected in the San Francisco Bay Area with a front-facing camera mounted on a vehicle. 

• This dataset focuses on road scene understanding with ephasis on construction zones - a construction zone exists in most of the video clips in the dataset.

• The dataset has perspective image annotations (topology related annotations, planning related annotations, affordance annotations, lane marking attributes). Examples of topology related annotations & planning related annotations are shown in the Dataset Visualization section below. There are about 5623 images with such annotations in the dataset.

• The dataset has Bird'e Eye View annotations (topology related annotations, planning related annotations). Examples of topology related annotations & planning related annotations are shown in the Dataset Visualization section below. There are about 87503 images with such annotations in the dataset.