An Empirical Study of Propagation-based Methods for Video Object Segmentation
This work provides a standardized benchmark for researchers in video object segmentation, though it is incremental as it builds on existing propagation-based methods.
The paper tackled the lack of fair comparisons in propagation-based video object segmentation by conducting an empirical study with unified settings, resulting in improved end-to-end memory networks achieving a global mean of 76.1 on the DAVIS 2017 val set.
While propagation-based approaches have achieved state-of-the-art performance for video object segmentation, the literature lacks a fair comparison of different methods using the same settings. In this paper, we carry out an empirical study for propagation-based methods. We view these approaches from a unified perspective and conduct detailed ablation study for core methods, input cues, multi-object combination and training strategies. With careful designs, our improved end-to-end memory networks achieve a global mean of 76.1 on DAVIS 2017 val set.