Finding Near-Duplicate Videos in Large-Scale Collections

Abstract

This chapter discusses the problem of Near-Duplicate Video Retrieval (NDVR). The main objective of a typical NDVR approach is: given a query video, retrieve all near-duplicate videos in a video repository and rank them based on their similarity to the query. Several approaches have been introduced in the literature, which can be roughly classified in three categories based on the level of video matching, i.e. (i) video-level, (ii) frame-level and (iii) filter-and-refine matching. Two methods based on video-level matching are presented in this chapter. The first is an unsupervised scheme that relies on a modified Bag-of-Word (BoW) video representation. The second is a supervised method based on Deep Metric Learning (DML). For the development of both methods, features are extracted from the intermediate layers of Convolutional Neural Networks and leveraged as frame descriptors, since they offer a compact and informative image representation, and lead to increased system efficiency. Extensive evaluation has been conducted on publicly available benchmark datasets, and the presented methods are compared with state-of-art approaches, achieving the best results in all evaluation setups.

Image Deduplication Using Efficient Visual Indexing and Retrieval: Optimizing Storage, Time and Energy for Deep Neural Network Training

“…For instance, Kaur et al [2] proposed a CNN-based online image deduplication technique, showing superior performance in detecting exact and nearexact images. Kordopatis-Zilos et al [19] introduced a method for near-duplicate video retrieval using unsupervised and supervised approaches based on Deep Metric Learning (DML). Liang et al [20] presented a hierarchical detection method utilizing CNN models and semantic features for near-duplicate video identiőcation.…”

Challenges for Automatic Detection of Fake News Related to Migration : Invited paper

“…[17] address the scalability issue by inferring possible provenance relations from metadata and thus significantly reduce the number of content matches that need to be performed. For video, this problem (also known as nearduplicate video retrieval) is even more computationally demanding [18]. In order to facilitate scalable matching of video content, approaches for compact descriptors (based on both hand-crafted and learned features) have been proposed.…”