FLIC-plus Dataset

Overview:

The FLIC-plus dataset is a subset of the FLIC-full training set provided by Sapp et al. It was first introduced as part of our NIPS 2014 paper: Joint Training of a Convolutional Network and a Graphical Model for Human Pose Estimation, for fair evaluation on the original FLIC test set.

For motivation and details on how the dataset was created, please refer to our paper.

To Download:

  • Download FLIC-full.zip hosted by Sapp et al. here. IF THE DATASET LINK IS BROKEN, CONTACT SAPP ET AL. WE DO NOT HOST IT.
  • Download our index list tr_plus_indices.mat
  • Load examples.mat (after decompressing FLIC-full.zip) and tr_plus_indices.mat into Matlab. The struct examples contains the image annotations for the 20928 FLIC-full training set. The double array tr_plus_indices contains the indices into the examples struct of the 17380 FLIC-full training set examples. Therefore, the first FLIC-plus anntation will be examples(tr_plus_indices(1)), the second examples(tr_plus_indices(2)), etc

Citing our paper (and this dataset):

@article{tompson14nips,
  author = {Jonathan Tompson and Arjun Jain and Yann Lecun and Christoph Bregler}
  title = {Joint Training of a Convolutional Network and a Graphical Model for Human Pose Estimation,
  journal = {NIPS},
  year = {2014},
}

Please make sure you also cite Sapp et al. for first introducing the FLIC dataset!

Publication

NIPS'14 paper