Rethinking Space-Time Networks with Improved Memory Coverage for Efficient Video Object Segmentation
arXiv:2106.05210
Abstract
This paper presents a simple yet effective approach to modeling space-time correspondences in the context of video object segmentation. Unlike most existing approaches, we establish correspondences directly between frames without re-encoding the mask features for every object, leading to a highly efficient and robust framework. With the correspondences, every node in the current query frame is inferred by aggregating features from the past in an associative fashion. We cast the aggregation process as a voting problem and find that the existing inner-product affinity leads to poor use of memory with a small (fixed) subset of memory nodes dominating the votes, regardless of the query. In light of this phenomenon, we propose using the negative squared Euclidean distance instead to compute the affinities. We validated that every memory node now has a chance to contribute, and experimentally showed that such diversified voting is beneficial to both memory efficiency and inference accuracy. The synergy of correspondence networks and diversified voting works exceedingly well, achieves new state-of-the-art results on both DAVIS and YouTubeVOS datasets while running significantly faster at 20+ FPS for multiple objects without bells and whistles.
Accepted to NeurIPS 2021. Project page: https://hkchengrex.github.io/STCN/
References in corpus (9)
- Rethinking Atrous Convolution for Semantic Image Segmentation
- High-Speed Tracking with Kernelized Correlation Filters
- Do Convnets Learn Correspondence?
- FEELVOS: Fast End-to-End Embedding Learning for Video Object Segmentation
- CascadePSP: Toward Class-Agnostic and Very High-Resolution Segmentation via Global and Local Refinement
- State-Aware Tracker for Real-Time Video Object Segmentation
- Spatiotemporal CNN for Video Object Segmentation
- Learning Position and Target Consistency for Memory-based Video Object Segmentation
- Fast Video Object Segmentation With Temporal Aggregation Network and Dynamic Template Matching