Video Compression Using Deep Learning Approach on Drone Video Footage

Dini Adni Navastara*, Reza Adipatria Maranatha, Ary Mazharuddin Shiddiqi

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper implements a video compression system using a deep learning approach based on autoencoder architecture. This method is an end-to-end video compression model that combines motion estimation, motion compression, motion compensation, residual compression, and entropy encoding. It is objective to compress drone video footage into a more feasible form for further usage. The comparison between the proposed model and traditional compression algorithm on the test scenarios, such as H.264 and H.265 (HEVC), is evaluated using the metric of PSNR and MS-SSIM. Based on the experimental result, the highest performance of the proposed model on the MS-SSIM metric is yielded on the UVG-Beauty dataset with an MS-SSIM score of 0.943, λ=64, and BPP value of 0.452. While, the highest performance of the proposed model on PSNR metric is obtained on the drone video data with a PSNR score of 36.88 dB, λ=2048, and BPP value of 0.206.

Original languageEnglish
Title of host publicationProceedings of the International Conference on Advanced Technology and Multidiscipline, ICATAM 2021
Subtitle of host publication"Advanced Technology and Multidisciplinary Prospective Towards Bright Future" Faculty of Advanced Technology and Multidiscipline
EditorsPrihartini Widiyanti, Prastika Krisma Jiwanti, Gunawan Setia Prihandana, Ratih Ardiati Ningrum, Rizki Putra Prastio, Herlambang Setiadi, Intan Nurul Rizki
PublisherAmerican Institute of Physics Inc.
ISBN (Electronic)9780735444423
DOIs
Publication statusPublished - 19 May 2023
Event1st International Conference on Advanced Technology and Multidiscipline: Advanced Technology and Multidisciplinary Prospective Towards Bright Future, ICATAM 2021 - Virtual, Online
Duration: 13 Oct 202114 Oct 2021

Publication series

NameAIP Conference Proceedings
Volume2536
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Conference

Conference1st International Conference on Advanced Technology and Multidiscipline: Advanced Technology and Multidisciplinary Prospective Towards Bright Future, ICATAM 2021
CityVirtual, Online
Period13/10/2114/10/21

Fingerprint

Dive into the research topics of 'Video Compression Using Deep Learning Approach on Drone Video Footage'. Together they form a unique fingerprint.

Cite this