This article explains how Kuaishou’s UniDex replaces traditional term‑based inverted indexes with model‑driven semantic posting lists and how the end‑to‑end UniSearch framework generates video IDs directly from queries, delivering higher relevance, lower latency, and significant online performance gains.
This article presents Ant Group's multimodal research on video retrieval, detailing a large Chinese video‑text pre‑training dataset, three techniques that raise video‑text semantic search performance by up to 24.5%, and an end‑to‑end video‑video copyright detection system that reduces storage by 85% and speeds up inference 18‑fold.
This article presents Ant Group's multimodal research on video retrieval, detailing video‑text semantic search and video‑video same‑source search, introducing a large Chinese pre‑training dataset, novel pre‑training, hard‑sample mining, fine‑grained modeling techniques, and an efficient end‑to‑end copyright detection framework.
This article explores how modern video AI techniques—ranging from transformer‑based classification to semi‑supervised retrieval and token‑halting acceleration—enable efficient, accurate detection of prohibited content and fast, scalable video search in the era of short‑form media.
This article presents concise overviews of five AAAI 2024 accepted papers covering multi‑stage reinforcement‑learning recommendation, error‑adaptive watch‑time prediction, coarse‑to‑fine text‑to‑video retrieval, enhanced fashion image retrieval, and conditional image‑to‑video generation, each with authors, download links, and reported performance gains.
This article presents a multimodal approach that combines dynamic analysis and graph machine learning to generate and apply social relationship graphs in videos, detailing problem background, graph generation modules, applications such as video retrieval, experimental results, and future research directions.
Tencent’s newly unveiled HunYuan AI model achieved a grand‑slam by ranking first on the five most authoritative cross‑modal video retrieval datasets, showcasing a hierarchical multimodal approach that dramatically boosts retrieval precision and promises broad impact for both research and industry applications.
Youku’s new text‑video alignment system automatically generates short video summaries by extracting multimodal video and linguistic features, matching sentences to clips through embedding and tag‑level models, and enabling AI‑driven auto‑editing that cuts production time from days to minutes.
An advanced multi‑level, multi‑modal search and graph engine for Youku processes text, voice, image and video queries across hierarchical video elements, using combined vector and inverted indexes to merge cross‑level and cross‑modal results, while a distributed knowledge‑graph layer enables multimodal graph traversal for billion‑scale video retrieval.
This article details a two‑stage video copyright detection pipeline that builds a frame‑level feature library, uses Hessian‑Affine + SIFT and Fisher Vectors for robust feature extraction, applies weighted bipartite graph matching and longest increasing subsequence localization, and achieves an F1‑score of 0.9086 on the CCF 2019 competition dataset.
The BoYun Vision team developed a CDVS‑ and CDVA‑based video copyright detection system that extracts key frames, combines handcrafted and deep features, aligns timestamps with a CDVS temporal node filter, and achieved a top‑2 result in the 2019 iQIYI‑CCF competition.
The Zheey team from Beijing University of Posts and Telecommunications tackled the iQIYI 2019 Multimodal Video Person Recognition Challenge with a three‑layer MLP on official face features, boosting a baseline 0.8742 to 0.8949 through model fusion, quality filtering and fine‑tuning, ultimately ranking sixth and open‑sourcing their code.
The 2018 Suning Search & Recommendation Technology Conference in Nanjing gathered over 400 industry experts to discuss search engine evolution, recommendation algorithm models, multi‑source data fusion, multimedia video retrieval, and AI‑driven advertising, highlighting practical implementations and future research directions.
