CapsuleVOS: Semi-Supervised Video Object Segmentation Using Capsule Routing

Explore semi-supervised video object segmentation using capsule routing in this 49-minute lecture from the University of Central Florida. Delve into capsule networks, computer graphics, and inverse graphics before examining different capsule formulations and the concept of capsule routing. Learn about video capsule networks and their application to semi-supervised video object segmentation. Analyze datasets, including examples from DAVIS, and understand the architecture of CapsuleVOS, including its video encoder, frame encoder with memory module, attention routing, and convolutional capsule layer. Evaluate quantitative results from the YoutubeVOS dataset and speed analysis, as well as qualitative results for single and multiple objects. Examine the effects of memory and zooming modules on the system's performance.