YouTube-VIS 2021

Emerging

4papers using it

2022first seen

YouTube-VIS 2021 is a benchmark dataset for Video Instance Segmentation that contains video sequences annotated with object instance masks, used to evaluate the performance of VIS methods in accurately associating and segmenting objects across frames.

🔎 Find this dataset

Papers using YouTube-VIS 2021 (4)

FlowCut: Unsupervised Video Instance Segmentation via Temporal Mask Matching2025

Temporally Efficient Vision Transformer for Video Instance Segmentation2022 · 2 cites

Consistent Video Instance Segmentation with Inter-Frame Recurrent Attention2022 · 1 cites

VISAGE: Video Instance Segmentation with Appearance-Guided Enhancement2023 · 1 cites