Introduction

Most video clips are between 15 − 60 sec long and are recorded around construction zones. For each video about 10 frames are manually selected and annotated. We annotate a total of 5, 630 perspective images.

Using COLMAP, we reconstruct a 3D dense point cloud given a video clip. We then annotate semantic labels of 3D points manually. A total of about 70, 000 BEV  image / ground truth pairs are constructed.

The below showing distribution of classes and corresponding attributes in the perspective EPOSH dataset. The left subplot shows classes and the right subplot shows the corresponding attributes and affordance classes in the dataset.

Dataset Visualization

Data Format

DATA:

pers                                              # perspective annotations
   |
   video_id
           |
           img/frame_n.png                        # images
           viz-img                                # color coded annotations
           label-topo/frame_n.png                 # topology related annotations
           label-plan/frame_n.png                 # planning related annotations
           label-lane-markings/frame_n.png        # lane marking annotations
           label-affordance/frame_n.png           # affordance annotations

bev                                               # bev annotations
   |
   video_id
           |
           rectified_resized_img/frame_n.png      # rectified resized images
           rectified_resized_img-viz/frame_n.png  # color coded annotations
           label-topo/frame_n.png                 # topology related annotations
           label-plan/frame_n.png                 # planning related annotations
           resized_rectified_vid.mp4              # rectified resized video
           vid.mp4                                # original resized video
vids
   |
   video_id.mp4                                   # set of videos

splits
   |
   bev_train.txt
   bev_val.txt
   pers_train.txt
   pers_val.txt
  

Please refer to the dataset readme for more details.

Download the dataset

The dataset is available for non-commercial usage. You must be affiliated with a university and use your university email address to make the request. Use this link to make the download request.

Citation

This dataset corresponds to the paper, Bird’s Eye View Segmentation Using Lifted 2D Semantic Features", as it appears in the proceedings of British Machine Vision Conference (BMVC) 2021".  In the current release, the data is available for researchers from universities.

Please cite the following paper if you find the dataset useful in your work:

@inproceedings{dwivedi2021bird,
  title={Bird’s eye view segmentation using lifted 2D semantic features},
  author={Dwivedi, Isht and Malla, Srikanth and Chen, Yi-Ting and Dariush, Behzad},
  booktitle={British Machine Vision Conference (BMVC)},
  pages={6985--6994},
  year={2021}
}