August 2024
Intermediate to advanced
370 pages
9h 45m
English
Our objective in this chapter is summarizing first-person (egocentric) videos at frame level. Two different solutions are proposed. The first solution explores different graph representations, which include graph-based shot boundary detection, graph-based center-surround model generation and feature extraction, followed by a modified Minimum Spanning Tree (MST) based clustering. We show how the above components can be combined to achieve accurate yet computationally efficient egocentric video summarization. We present a second solution for egocentric video summarization using deep features and an optimal clustering approach. Based on an augmented pretrained convolutional neural network (CNN), ...
Read now
Unlock full access