With the rise of "Metaverse" and "Web 3.0", Non-Fungible Token (NFT) has emerged as a kind of pivotal digital asset, garnering significant attention. By the end of March 2024, more than 1.7 billion NFTs have been minted across various blockchain platforms. To effectively locate a desired NFT, conducting searches within a vast array of NFTs is essential. The challenge in NFT retrieval is heightened due to the high degree of similarity among different NFTs, regarding regional and semantic aspects. In this paper, we will introduce a benchmark dataset named "NFT Top1000 Visual-Text Dataset" (NFT1000), containing 7.56 million image-text pairs, and being collected from 1000 most famous PFP1 NFT collections2 by sales volume on the Ethereum blockchain. Based on this dataset and leveraging the CLIP series of pre-trained models as our foundation, we propose the dynamic masking fine-tuning scheme. This innovative approach results in a 7.4\% improvement in the top1 accuracy rate, while utilizing merely 13\% of the total training data (0.79 million vs. 6.1 million). We also propose a robust metric Comprehensive Variance Index (CVI) to assess the similarity and retrieval difficulty of visual-text pairs data. The dataset will be released as an open-source resource. For more details, please refer to: https://github.com/ShuxunoO/NFT-Net.git.
Evaluating1 the popularity of content, developing personalized recommendation algorithms, and optimizing advertising revenue are crucial aspects of Over-the-Top (OTT) media content services. In order to achieve these goals, accurate viewership measurement of OTT content is essential. However, due to the stringent security policies of mobile platforms, it has been challenging to propose an easy and reliable method for measuring viewership. In this paper, we propose an efficient and secure viewership measurement technique for OTT content on the iOS platforms. The proposed technique utilizes audio filtering and zero-knowledge proof techniques to address the challenges. By leveraging audio filtering, the content can be quickly identified, enabling accurate viewership measurement. Additionally, the use of zero-knowledge proof ensures the protection of users' personal information, preventing indiscriminate acquisition of user data in OTT services.
The last decades have seen video production and consumption rise significantly: TV/cinematography, social networking, digital marketing, and video surveillance incrementally and cumulatively turned video content into the predilection type of data to be exchanged, stored, and processed. Belonging to video processing realm, video fingerprinting (also referred to as content-based copy detection or near duplicate detection ) regroups research efforts devoted to identifying duplicated and/or replicated versions of a given video sequence (query) in a reference video dataset. The present paper reports on a state-of-the-art study on the past and present of video fingerprinting, while attempting to identify trends for its development. First, the conceptual basis and evaluation frameworks are set. This way, the methodological approaches (situated at the cross-roads of image processing, machine learning, and neural networks) can be structured and discussed. Finally, fingerprinting is confronted to the challenges raised by the emerging video applications ( e.g. , unmanned vehicles or fake news) and to the constraints they set in terms of content traceability and computational complexity. The relationship with other technologies for content tracking ( e.g., DLT - Distributed Ledger Technologies) are also presented and discussed.
Open access
Advanced Steganography and Watermarking Techniques