MAIN: Multi-Attention Instance Network for video segmentation

Instance-level video segmentation requires a solid integration of spatial and temporal information. However, current methods rely mostly on domain-specific information (online learning) to produce accurate instance-level segmentations. We propose novel approach that relies exclusively the generic spatio-temporal attention cues. Our strategy, named Multi-Attention Instance Network (MAIN), overcomes challenging scenarios over arbitrary videos without modeling sequence- or instance-specific knowledge. design MAIN segment multiple instances in single forward pass, optimize it with loss function favors class agnostic predictions assigns penalties. achieve state-of-the-art performance Youtube-VOS dataset benchmark, improving unseen Jaccard F-Metric by 6.8% 12.7% respectively, while operating at real-time (30.3 FPS).