High-quality Activity-Level Video Advertising

Yuan, Mu; Zhang, Lan; Wu, Zhengtao; Zheng, Daren

doi:10.1109/iwqos49365.2020.9212956

Cited by 5 publications

(3 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These authors matched ad content by fusing semantic, scene, emotion, object, audio, and color feature information. Yuan et al (2020) carried out advertisement matching from five perspectives, including facial features, item types, scene categories, human behaviors, and video subtitles. They also used a heuristic algorithm based on non‐maximum suppression to ensure the intrusiveness to the viewers is under a certain constraint.…”

Section: Computer Science Studiesmentioning

confidence: 99%

A survey of online video advertising

Zhang

Yan

et al. 2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

With the development of social media and the ubiquity of the Internet, recent years have witnessed the rapid development of online video advertising among publishers and advertisers. Video advertising, as a new type of advertisement, has gained significant research attention from both academia and industry, coinciding with the ever-growing volume of online videos. In this research, we provide a comprehensive survey of online video advertising in the fields of social science and computer science. We investigate state-of-the-art articles from 1990 to the present and provide a new taxonomy of extant research topics based on these articles. We also highlight the factors that cause advertising to affect people and the most popular video advertising techniques used in computer science. Finally, on the basis of the analytics of the surveyed papers, future challenges are identified and potential solutions to these are discussed.

show abstract

Section: Computer Science Studiesmentioning

confidence: 99%

A survey of online video advertising

Zhang

Yan

et al. 2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

show abstract

“…We only assume that input spaces {X i } k i=1 are the same or aligned. The case that different models share the same input spaces is common, e.g., multi-task learning-based robotics (Crawshaw 2020;Zhang and Yang 2021) and multimedia advertising (Yuan et al 2020b). The aligned input spaces typically exists in the context of multimodal scenarios, e.g., multi-modality event detection (Elhoseiny et al 2016) and visual speech synthesis (Baltrušaitis, Ahuja, and Morency 2018).…”

Section: Problem Statementmentioning

confidence: 99%

MLink: Linking Black-Box Models for Collaborative Multi-Model Inference

Yuan

Zhang

2022

AAAI

Self Cite

View full text Add to dashboard Cite

The cost efficiency of model inference is critical to real-world machine learning (ML) applications, especially for delay-sensitive tasks and resource-limited devices. A typical dilemma is: in order to provide complex intelligent services (e.g. smart city), we need inference results of multiple ML models, but the cost budget (e.g. GPU memory) is not enough to run all of them. In this work, we study underlying relationships among black-box ML models and propose a novel learning task: model linking. Model linking aims to bridge the knowledge of different black-box models by learning mappings (dubbed model links) between their output spaces. Based on model links, we developed a scheduling algorithm, named MLink. Through collaborative multi-model inference enabled by model links, MLink can improve the accuracy of obtained inference results under the cost budget. We evaluated MLink on a multi-modal dataset with seven different ML models and two real-world video analytics systems with six ML models and 3,264 hours of video. Experimental results show that our proposed model links can be effectively built among various black-box models. Under the budget of GPU memory, MLink can save 66.7% inference computations while preserving 94% inference accuracy, which outperforms multi-task learning, deep reinforcement learning-based scheduler and frame filtering baselines.

show abstract

“…We only assume that input spaces {X i } k i=1 are the same or aligned. The case that different models share the same input spaces is common, e.g., multi-task learningbased robotics [8], [9] and multimedia advertising [23]. The aligned input spaces typically exists in the context of multimodal scenarios, e.g., multi-modality event detection [24] and visual speech synthesis [25].…”

Section: Problem Statementmentioning

confidence: 99%

MLink: Linking Black-Box Models from Multiple Domains for Collaborative Inference

Yuan¹,

Zhang²,

Zheng³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

The cost efficiency of model inference is critical to real-world machine learning (ML) applications, especially for delay-sensitive tasks and resource-limited devices. A typical dilemma is: in order to provide complex intelligent services (e.g. smart city), we need inference results of multiple ML models, but the cost budget (e.g. GPU memory) is not enough to run all of them. In this work, we study underlying relationships among black-box ML models and propose a novel learning task: model linking, which aims to bridge the knowledge of different black-box models by learning mappings (dubbed model links) between their output spaces. We propose the design of model links which supports linking heterogeneous black-box ML models. Also, in order to address the distribution discrepancy challenge, we present adaptation and aggregation methods of model links. Based on our proposed model links, we developed a scheduling algorithm, named MLink. Through collaborative multi-model inference enabled by model links, MLink can improve the accuracy of obtained inference results under the cost budget. We evaluated MLink on a multi-modal dataset with seven different ML models and two real-world video analytics systems with six ML models and 3,264 hours of video. Experimental results show that our proposed model links can be effectively built among various black-box models. Under the budget of GPU memory, MLink can save 66.7% inference computations while preserving 94% inference accuracy, which outperforms multi-task learning, deep reinforcement learning-based scheduler and frame filtering baselines.

show abstract

High-quality Activity-Level Video Advertising

Cited by 5 publications

References 19 publications

A survey of online video advertising

A survey of online video advertising

MLink: Linking Black-Box Models for Collaborative Multi-Model Inference

MLink: Linking Black-Box Models from Multiple Domains for Collaborative Inference

Contact Info

Product

Resources

About