Rethinking Space-Time Networks with Improved Memory Coverage for Efficient Video Object Segmentation

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.

arXiv commentAccepted to NeurIPS 2021. Project page: https://hkchengrex.github.io/STCN/

References in corpus (19)

• Rethinking Atrous Convolution for Semantic Image Segmentation
• High-Speed Tracking with Kernelized Correlation Filters
• Prototypical Networks for Few-shot Learning
• MOT16: A Benchmark for Multi-Object Tracking
• YouTube-VOS: A Large-Scale Video Object Segmentation Benchmark
• RANet: Ranking Attention Network for Fast Video Object Segmentation
• FSS-1000: A 1000-Class Dataset for Few-Shot Segmentation
• Do Convnets Learn Correspondence?
• Efficient Video Object Segmentation via Network Modulation
• The Lipschitz Constant of Self-Attention
• FEELVOS: Fast End-to-End Embedding Learning for Video Object Segmentation
• SCNet: Learning Semantic Correspondence
• Blazingly Fast Video Object Segmentation with Pixel-Wise Metric Learning
• CascadePSP: Toward Class-Agnostic and Very High-Resolution Segmentation via Global and Local Refinement
• State-Aware Tracker for Real-Time Video Object Segmentation
• Efficient Regional Memory Network for 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

Cited by in corpus (10)

• Fully Transformer-Equipped Architecture for End-to-End Referring Video Object Segmentation
• HODOR: High-level Object Descriptors for Object Re-segmentation in Video Learned from Static Images
• ViDSOD-100: A New Dataset and a Baseline Model for RGB-D Video Salient Object Detection
• Timeline and Boundary Guided Diffusion Network for Video Shadow Detection
• Neighbor Correspondence Matching for Flow-based Video Frame Synthesis
• Learning Motion and Temporal Cues for Unsupervised Video Object Segmentation
• SpVOS: Efficient Video Object Segmentation with Triple Sparse Convolution
• UVOSAM: A Mask-free Paradigm for Unsupervised Video Object Segmentation via Segment Anything Model
• FiVOS: A Fish Segmentation Algorithm Based on Interactive Video Object Segmentation and Filter Enhancement
• Enhancing Self-Supervised Fine-Grained Video Object Tracking with Dynamic Memory Prediction