Benchmark Video Object Segmentation on DAVIS 2016

Quantitative state-of-the-art comparison

Update 16-07-2020:
We no longer maintain these leaderbaords. You can find the latest results in the following websites:

  • Semi-supervised: val
  • Unsupervised: val

Select the type of table:
Select the dataset subset*:
*Note about the subsets: A dash "-" means that a certain technique used the training set to train the algorithm and so only results on "val" make sense. In the future and in the challenge, results in the validation/test will be used, but we keep the "all" evaluation to be coherent with the original paper.
Semi-Supervised techniques are depicted in red and Unsupervised in green.


Is your technique missing although it's published and the code is public? Let us know and we'll add it.