Short Video Plagiarism Detection Based on Deep Features and Shallow Metadata

Fengnian Zhao

doi:10.47297/wspciWSP2516-252713.20261004

PDF
Export
Share
Collection
Album

Short Video Plagiarism Detection Based on Deep Features and Shallow Metadata

Volume 10, Issue 4

Views: 0 Downloads: 0

Short Video Plagiarism Detection Based on Deep Features and Shallow Metadata
Short Video Plagiarism Detection Based on Deep Features and Shallow Metadata
Leicester International Institute, Dalian University of Technology, Dalian, Liaoning, 116024, China
Vol. 10, Issue 4, Pages: 65-68(2026)
Published：30 August 2026
DOI：10.47297/wspciWSP2516-252713.20261004
Accepted：

Scan QR Code

Zhao Fengnian.Short Video Plagiarism Detection Based on Deep Features and Shallow Metadata[J].Creativity and Innovation,2026,10(04):65-68.
DOI：

Zhao Fengnian.Short Video Plagiarism Detection Based on Deep Features and Shallow Metadata[J].Creativity and Innovation,2026,10(04):65-68. DOI： 10.47297/wspciWSP2516-252713.20261004.

摘要

Abstract

Short-video platforms face a growing problem of content reposting

and plagiarism has evolved from whole-video copying to complex forms such as segment extraction

mirroring

and speed changes. This paper addresses this scenario by building a video-pair-level plagiarism detection method: shallow metadata such as resolution

bitrate

and frame rate are extracted for each video

and a pre-trained ResNet50 is used to extract deep features from 10 sampled frames. The features of the two videos are combined into a pair-level representation

on which a random forest classifier is trained to determine whether plagiarism exists. On 528 videos from the VCDB dataset

with the training and test sets split by topic

the method achieves an accuracy of 91.2% and a ROC-AUC of 0.974 on the test set. The experiments further reveal that the maximum similarity in the 10×10 frame-alignment matrix is the strongest discriminative signal: using a threshold of 0.60 alone yields a recall of 98.1%. The proposed method is lightweight and deployable

making it suitable for coarse screening and copyright monitoring on short-video platforms

although its accuracy on strong transformations such as mirroring and cropping remains limited.

关键词

Keywords

references

Sivic, & Zisserman. (2003). Video google: a text retrieval approach to object matching in videos. IEEE.

Jegou, H. , Douze, M. , & Schmid, C. . (2008). Hamming embedding and weak geometric consistency for large scale image search. Springer-Verlag.

Douze, M. , Jegou, H. , & Schmid, C. . (2010). An image-based approach to video copy detection with spatio-temporal post-filtering. IEEE Transactions on Multimedia, 12(4), 257-266.

Liong, V. E. , Lu, J. , Wang, G. , Moulin, P. , & Zhou, J. . (2015). Deep hashing for compact binary codes learning. IEEE.

Wu, X. , Hauptmann, A. G. , & Ngo, C. . (2007). Practical elimination of near-duplicates from web video search.

Jiang, Y. G. , Jiang, Y. , & Wang, J. . (2014). VCDB: A Large-Scale Database for Partial Copy Detection in Videos. European Conference on Computer Vision. Springer, Cham.

Krizhevsky, A. , Sutskever, I. , & Hinton, G. . (2012). ImageNet Classification with Deep Convolutional Neural Networks. NIPS (Vol.25). Curran Associates Inc.

Simonyan, K. , & Zisserman, A. . (2014). Very deep convolutional networks for large-scale image recognition. arXiv e-prints.

He, K. , Zhang, X. , Ren, S. , & Sun, J. . (2016). Deep residual learning for image recognition. IEEE.

Deng, J. , Dong, W. , Socher, R. , Li, L. J. , Li, K. , & Fei-Fei, L. . (2009). Imagenet: a large-scale hierarchical image database. Proc of IEEE Computer Vision & Pattern Recognition, 248-255.