A Multi-Scale Recurrent Framework for Motion Segmentation with Event Camera

Motion segmentation is a formidable computer vision task, aiming to segment moving targets from dynamic scene. In this paper, we choose introduce an additional modality bolster the robustness. The event camera bio-inspired sensor that accurately detects and captures intensity changes with exceptional temporal resolution range, which optimal choice for motion segmentation. Therefore, present novel framework event-based propose Multi-Scale Recurrent Neural Network (MSRNN) fuse information efficiently. To our best knowledge, it first time multi-scale recurrent architecture implemented in proposed evaluated through experiments conducted on EV-IMO dataset. Our method achieves mean Intersection-over-Union (mIoU) of 82.0%, sets new state-of-the-art further validate approach arduous real-world scenarios, Event Challenging dataset, consisting 350 images corresponding events, outperforms other methods by 1.5% (IoU).