Multi-Granularity Video Object Segmentation
This addresses the gap in video segmentation benchmarks for real-world scenarios by providing a dataset that includes non-salient objects, which is incremental as it builds on existing segmentation methods.
The authors tackled the problem of video segmentation being limited to salient objects by creating a new dataset (MUG-VOS) with multi-granularity annotations for both salient and non-salient masks, and their proposed memory-based mask propagation model achieved the best performance among existing methods.
Current benchmarks for video segmentation are limited to annotating only salient objects (i.e., foreground instances). Despite their impressive architectural designs, previous works trained on these benchmarks have struggled to adapt to real-world scenarios. Thus, developing a new video segmentation dataset aimed at tracking multi-granularity segmentation target in the video scene is necessary. In this work, we aim to generate multi-granularity video segmentation dataset that is annotated for both salient and non-salient masks. To achieve this, we propose a large-scale, densely annotated multi-granularity video object segmentation (MUG-VOS) dataset that includes various types and granularities of mask annotations. We automatically collected a training set that assists in tracking both salient and non-salient objects, and we also curated a human-annotated test set for reliable evaluation. In addition, we present memory-based mask propagation model (MMPM), trained and evaluated on MUG-VOS dataset, which leads to the best performance among the existing video object segmentation methods and Segment SAM-based video segmentation methods. Project page is available at https://cvlab-kaist.github.io/MUG-VOS.