The "Flying Chairs" are a synthetic dataset with optical flow ground truth. It consists of 22872 image pairs and corresponding flow fields. Images show renderings of 3D chair models moving in front of random backgrounds from Flickr. Motions of both the chairs and the background are purely planar.
This dataset has been used for training convolutional networks in the ICCV 2015 paper FlowNet: Learning Optical Flow with Convolutional Networks.
The dataset is provided for research purposes only and without any warranty. Any commercial use is prohibited. If you use the dataset in your research, you should cite the aforementioned paper.
@InProceedings{DFIB15,
author = "A. Dosovitskiy and P. Fischer and E. Ilg and P. H{\"a}usser and C. Haz{\i}rba{\c{s}} and V. Golkov and P. v.d. Smagt and D. Cremers and T. Brox",
title = "FlowNet: Learning Optical Flow with Convolutional Networks",
booktitle = "IEEE International Conference on Computer Vision (ICCV)",
month = " ",
year = "2015",
url = "http://lmb.informatik.uni-freiburg.de/Publications/2015/DFIB15"
}
Download the "Flying Chairs" dataset
"Flying chairs" train-validation split (1 - train, 2 - validation)
Sintel train-validation split (1 - train, 2 - validation)
Download handy Python IO routines
The Flying Chairs 2 dataset is generated using the same settings as the Flying Chairs dataset, but contains additional modalities that were used to train the networks in the paper Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation.
All modalities provided are:
The dataset is provided for research purposes only and without any warranty. Any commercial use is prohibited. If you use the dataset in your research, you should cite the following two papers:
@InProceedings{DFIB15,
author = "A. Dosovitskiy and P. Fischer and E. Ilg and P. H{\"a}usser and C. Haz{\i}rba{\c{s}} and V. Golkov and P. v.d. Smagt and D. Cremers and T. Brox",
title = "FlowNet: Learning Optical Flow with Convolutional Networks",
booktitle = "IEEE International Conference on Computer Vision (ICCV)",
month = " ",
year = "2015",
url = "http://lmb.informatik.uni-freiburg.de/Publications/2015/DFIB15"
}
@InProceedings{ISKB18,
author = "E. Ilg and T. Saikia and M. Keuper and T. Brox",
title = "Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation",
booktitle = "European Conference on Computer Vision (ECCV)",
month = " ",
year = "2018",
url = "http://lmb.informatik.uni-freiburg.de/Publications/2018/ISKB18"
}
Download the "Flying Chairs 2" dataset
Download handy Python IO routines
"ChairsSDHom" is a synthetic dataset with optical flow ground truth. Designed to be robust to untextured regions and to produce flow magnitude histograms close to those of the UCF101 dataset, ChairsSDHom is a good candidate for training if you want your optical flow method to work well on real-world data and generally rather small displacements. The dataset is abstract enough to not overfit to any realistic scenario, so even if you have specialized data, ChairsSDHom may still serve as good additional or pretraining data.
This dataset has been used for training convolutional networks in the CVPR 2017 paper FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks.
The dataset is provided for research purposes only and without any warranty. Any commercial use is prohibited. If you use the dataset in your research, you should cite the aforementioned paper.
@InProceedings{IMKDB17,
author = "E. Ilg and N. Mayer and T. Saikia and M. Keuper and A. Dosovitskiy and T. Brox",
title = "FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks",
booktitle = "IEEE Conference on Computer Vision and Pattern Recognition (CVPR)",
month = "Jul",
year = "2017",
url = "http://lmb.informatik.uni-freiburg.de//Publications/2017/IMKDB17"
}
Download the "ChairsSDHom" dataset
Download handy Python IO routines
The extended version contains the same flows and images, but also additional modalities that were used to train the networks in the paper Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation.
All modalities provided are:
The dataset is provided for research purposes only and without any warranty. Any commercial use is prohibited. If you use the dataset in your research, you should cite the following two papers:
@InProceedings{IMKDB17,
author = "E. Ilg and N. Mayer and T. Saikia and M. Keuper and A. Dosovitskiy and T. Brox",
title = "FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks",
booktitle = "IEEE Conference on Computer Vision and Pattern Recognition (CVPR)",
month = "Jul",
year = "2017",
url = "http://lmb.informatik.uni-freiburg.de//Publications/2017/IMKDB17"
}
@InProceedings{ISKB18,
author = "E. Ilg and T. Saikia and M. Keuper and T. Brox",
title = "Occlusions, Motion and Depth Boundaries with a Generic Network for Disparity, Optical Flow or Scene Flow Estimation",
booktitle = "European Conference on Computer Vision (ECCV)",
month = " ",
year = "2018",
url = "http://lmb.informatik.uni-freiburg.de/Publications/2018/ISKB18"
}
Download the "ChairsSDHom extended" dataset
Download handy Python IO routines