HighlightMe: Detecting Highlights from Human-Centric Videos

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bhattacharya, Uttaran, Wu, Gang, Petrangeli, Stefano, Swaminathan, Viswanathan, Manocha, Dinesh
Format: Preprint
Veröffentlicht: 2021
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912130749431808
author Bhattacharya, Uttaran
Wu, Gang
Petrangeli, Stefano
Swaminathan, Viswanathan
Manocha, Dinesh
author_facet Bhattacharya, Uttaran
Wu, Gang
Petrangeli, Stefano
Swaminathan, Viswanathan
Manocha, Dinesh
contents We present a domain- and user-preference-agnostic approach to detect highlightable excerpts from human-centric videos. Our method works on the graph-based representation of multiple observable human-centric modalities in the videos, such as poses and faces. We use an autoencoder network equipped with spatial-temporal graph convolutions to detect human activities and interactions based on these modalities. We train our network to map the activity- and interaction-based latent structural representations of the different modalities to per-frame highlight scores based on the representativeness of the frames. We use these scores to compute which frames to highlight and stitch contiguous frames to produce the excerpts. We train our network on the large-scale AVA-Kinetics action dataset and evaluate it on four benchmark video highlight datasets: DSH, TVSum, PHD2, and SumMe. We observe a 4-12% improvement in the mean average precision of matching the human-annotated highlights over state-of-the-art methods in these datasets, without requiring any user-provided preferences or dataset-specific fine-tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2110_01774
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle HighlightMe: Detecting Highlights from Human-Centric Videos
Bhattacharya, Uttaran
Wu, Gang
Petrangeli, Stefano
Swaminathan, Viswanathan
Manocha, Dinesh
Computer Vision and Pattern Recognition
We present a domain- and user-preference-agnostic approach to detect highlightable excerpts from human-centric videos. Our method works on the graph-based representation of multiple observable human-centric modalities in the videos, such as poses and faces. We use an autoencoder network equipped with spatial-temporal graph convolutions to detect human activities and interactions based on these modalities. We train our network to map the activity- and interaction-based latent structural representations of the different modalities to per-frame highlight scores based on the representativeness of the frames. We use these scores to compute which frames to highlight and stitch contiguous frames to produce the excerpts. We train our network on the large-scale AVA-Kinetics action dataset and evaluate it on four benchmark video highlight datasets: DSH, TVSum, PHD2, and SumMe. We observe a 4-12% improvement in the mean average precision of matching the human-annotated highlights over state-of-the-art methods in these datasets, without requiring any user-provided preferences or dataset-specific fine-tuning.
title HighlightMe: Detecting Highlights from Human-Centric Videos
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2110.01774