Long-term Video Object Segmentation
For each video in LVOS V2, we provide images, pixel-wise annotations, and attribute annotations. Images and annotations are at 6fps .
You can use the training set to train your own long-term VOS model. And we believe that the complex motion in LVOS is also helpful for short-term VOS. You can also try to train short-term VOS model on LVOS training set.
All the images and groundtruth masks are provided. Please use our evaluation tookit to assess your own algorithms.
Note: only images and the groundtruth masks are provided in testing set. The rest groundtruch masks are kept private. For V2 version, we do not have the plan to publish the evaluation server currently.
Download the meta jsons from the following links:
Google Drive | Baidu PanDownload the jsons with attributes from the following links:
Google Drive | Baidu PanFor each video in LVOS V1, we provide images, pixel-wise annotations, and lingual description. Images and annotations are at 6fps .
You can use the training set to train your own long-term VOS model. And we believe that the complex motion in LVOS is also helpful for short-term VOS. You can also try to train short-term VOS model on LVOS training set.
All the images and groundtruth masks are provided. Please use our evaluation tookit to assess your own algorithms.
Note: only images and the groundtruth masks are provided in testing set. The rest groundtruch masks are kept private. Please use the evaluation server for convenience of evaluating your own algorithms.
Download the meta jsons from the following links:
Google Drive | Kaggle | Baidu PanDownload the expression jsons from the following links:
Google Drive | Kaggle | Baidu PanIn LVOS, we utilize three metrcs: Region Similarity, Contour Accuracy, and Standard Deviation to measure the performance of VOS methods. The definitions of the three metrics can be seen in paper.
The evaluation toolkit for evaluation set can be found on Github. Please see the repository for more information.
The groundtruth masks of testing set are keep private. Please use the evaluation server instead.
We provide the tools and test scripts in this repository. Click on this link for more information.
Please consider citing LVOS if you use LVOS in your research.
# for LVOS V2
@article{hong2024lvos,
author = {Hong, Lingyi and Liu, Zhongying and Chen, Wenchao and Tan, Chenzhi and Feng, Yuang and Zhou, Xinyu and Guo, Pinxue and Li, Jinglun and Chen, Zhaoyu and Gao, Shuyong and others},
title = {LVOS: A Benchmark for Large-scale Long-term Video Object Segmentation},
journal = {arXiv preprint arXiv:2404.19326},
year = {2024},
}
# for LVOS V1
@InProceedings{Hong_2023_ICCV,
author = {Hong, Lingyi and Chen, Wenchao and Liu, Zhongying and Zhang, Wei and Guo, Pinxue and Chen, Zhaoyu and Zhang, Wenqiang},
title = {LVOS: A Benchmark for Long-term Video Object Segmentation},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
month = {October},
year = {2023},
pages = {13480-13492}
}