Skip to content

Benchmark Animal Dataset of Joint Annotations (BADJA) with example code, as introduced in "Creatures Great and SMAL: Recovering the shape and motion of animals from video" (ACCV 2018).

License

benjiebob/BADJA

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

BADJA

BADJA is the Benchmark Animal Dataset of Joint Annotations released with the paper Creatures Great and SMAL: Recovering the shape and motion of animals from video. If you use this dataset, please consider citing this work:

@inproceedings{biggs2018creatures,
  title={{C}reatures great and {SMAL}: {R}ecovering the shape and motion of animals from video},
  author={Biggs, Benjamin and Roddick, Thomas and Fitzgibbon, Andrew and Cipolla, Roberto},
  booktitle={ACCV},
  year={2018}
}

Annotations in are provided in this repository as *.json files for a total of 9 video sequences.

  1. Seven video sequences from DAVIS 2017 Video Segmentation dataset (Pont-Tuset et al. 2017),
  2. Two video sequences provided in extra_videos which were sourced by hand.

Note that raw frames for the tiger and cat sequences have been necessarily omitted from the extra_videos repository due to licensing restrictions.

We annotate 20 joints which are related to positions on the SMAL quadruped mesh (Zuffi et al. 2018) defined in `smal_CVPR2018.pkl'. This can be downloaded from the SMALR page.

The position of the first 16 joints (legs, neck, and tail) are given by SMAL model joints. The remainder (nose tip, chin, left ear and right ear) relate to specific SMAL vertices.

Position SMAL Vertex ID
Nose Tip 1863
Chin 26
Left Ear 149
Right Ear 2124

Annotations are provided approximately every 5 video frames with the exception of rs_dog which is annotated densely.

More detail on which SMAL joints have been annotated can be found in code/joint_catalog.py. The code folder also contains a Python script that displays annotated frames across the dataset.

Installation

  1. Clone the repository

    git clone https://github.com/benjiebob/BADJA.git
     
    
  2. Download DAVIS 2017 TrainVal Full-Resolution image and annotation frames and unpack to location "BADJA/DAVIS"

  3. Download Extra Videos and unpack to location `BADJA/extra_videos'

  4. Test everything is correctly located and working properly. Demo code requires numpy, scipy, opencv and matplotlib.

    python code/view_badja.py
    

Contribute

Please create a pull request or submit an issue if you would like to contribute.

About

Benchmark Animal Dataset of Joint Annotations (BADJA) with example code, as introduced in "Creatures Great and SMAL: Recovering the shape and motion of animals from video" (ACCV 2018).

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages