The LVIS dataset contains the same image data as the COCO dataset. For more details on that you can access our info page on the COCO dataset.
In particular, LVIS presents over 1200 categories and over 2 million high quality segmentation masks, as well as a broad selection of "rare categories".
For more information please read LVIS: A Dataset for Large Vocabulary Instance Segmentation.
Please follow this link for a comprehensive guide to the annotation methodology.
You can download the dataset here.
No official model has been provided for this dataset. However, you can consult models proposed by the contenters of the LVIS 2019 Challenge via the leaderboard here.
No official benchmark has been provided for this dataset. However, you can consult the LVIS 2019 Challenge leaderboard here.
Dataset licensed under the CC BY 4.0 license. All LVIS dataset images come from the COCO dataset.