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.

Overview
Introduction to Capsule Networks
Computer Graphics
Inverse Graphics
Different capsule formulations
What is capsule routing
Video Capsule Networks
Semi-Supervised Video Object Segmentation
Datasets
Example videos from DAVIS
VOS using Capsules
Video Encoder
Frame Encoder with Memory Module
Attention Routing
Conv Capsule Layer and Decoder Network
Objective Function
Quantitative Results - YoutubeVOS Dataset
Quantitative Results - Speed Analysis
Qualitative Results - Single Object
Qualitative Results - Multiple Objects
Effect of Memory Module
Effect of the Zooming Module
Effect of Zooming Module
CapsuleVOS Architecture