Multi-neural network based tiled 360°video caching with Mobile Edge Computing

Kumar, Shashwat; Bhagat, Lalit; Franklin, A. Antony; Jin, Jiong

doi:10.1016/j.jnca.2022.103342

Cited by 16 publications

(5 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This information is used to make informed caching decisions at the network edge, prioritizing popular content for enhanced network performance and user experience. The CNN architecture consists of multiple layers, each with a distinct purpose [44].…”

Section: Network Analysis For Trainingmentioning

confidence: 99%

“…These blocks play a crucial role in extracting a hierarchical structure of features, where basic patterns are detected at the early levels and more intricate patterns are identified as the input moves deeper into the network. By utilizing strides in the convolution operation, these blocks also decrease the dimensions of the feature maps, so enhancing the computational efficiency and reducing the complexity of the model [44], [45].…”

Section: Network Analysis For Trainingmentioning

confidence: 99%

“…MA360 is a system that uses multi-agent deep reinforcement learning to improve the quality of experience (QoE) and reduce bandwidth usage in live 360-degree video streaming. It addresses a more complex scenario compared to the previous focus on thumbnail optimization [44].…”

Section: Concatenationmentioning

confidence: 99%

See 2 more Smart Citations

Optimal Video Caching at The Edge of Network by Using Machine Learning

Ullah,

Xiaopeng,

Khan

2024

Preprint

View full text Add to dashboard Cite

Efficiently managing network resources in the dynamic field of video-on-demand (VoD) services is a significant challenge. This requires creative methods to optimize video caching strategies. This study examines the use of machine learning (ML), particularly reinforcement learning (RL), into edge network designs to improve video caching procedures. The decentralized, adaptable, and learning characteristics of machine learning offer a powerful method for effectively managing caching in real-time. This technique leads to reduced latency, optimized bandwidth utilization, and improved cache hit rates. Conventional caching algorithms frequently fail to adjust to the changing needs and popularity of material that is inherent in Video on Demand (VoD) services. Our study presents an innovative reinforcement learning (RL) caching method that is specifically developed to automatically adapt caching choices using real-time data analysis, thus guaranteeing an effective content delivery network. We created and evaluated three separate machine learning models: Convolutional Neural Networks (CNN) for extracting features, Deep Q-Networks (DQN) for generating decisions, and a hybrid model that combines the advantages of both CNN and DQN to achieve improved performance. By conducting thorough simulations and implementing our models in real-world deployment scenarios, we have shown substantial improvements compared to traditional caching methods in terms of scalability, adaptability to evolving user patterns, and overall network efficiency. This innovative method not only simplifies the transfer of content in Video on Demand (VoD) systems, but also establishes a new standard for integrating Artificial Intelligence (AI) technologies in network optimization. It effectively tackles present and future obstacles in the digital streaming industry.

show abstract

Section: Network Analysis For Trainingmentioning

confidence: 99%

Section: Network Analysis For Trainingmentioning

confidence: 99%

See 1 more Smart Citation

Optimal Video Caching at The Edge of Network by Using Machine Learning

Ullah,

Xiaopeng,

Khan

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Storage and computation resources are allocated jointly to improve cache hit rate and reduce latency and transmission cost. The authors in References 20–23 use similar solutions to optimize VR video transmission.…”

Section: Related Workmentioning

confidence: 99%

OJUMP: Optimization for joint unicast‐multicast panoramic VR live streaming system

Pan,

Zeng,

Zhuang

et al. 2023

Trans Emerging Tel Tech

View full text Add to dashboard Cite

Virtual reality (VR) videos have garnered increasing attention due to their ability to deliver immersive multimedia experiences to users. Nonetheless, the demanding requirements of high bandwidth and low latency for VR services present significant challenges, especially in the context of network infrastructure. In this article, we introduce a novel system called OJUMP, aimed at Optimizing the transmission and distribution of Joint Unicast‐Multicast Panoramic video field of view live streaming using Bit Index Explicit Replication (BIER). The proposed system encompasses a BIER‐based network transmission approach that enhances video transmission efficiency and facilitates large‐scale service deployment. Furthermore, OJUMP includes a dynamic disk‐state‐based tile slice distribution storage algorithm and a prefetch method designed for edge nodes to enhance the distribution efficiency of live streaming services. Experimental results illustrate that the proposed system offers distinct advantages over the baseline system in terms of bandwidth, latency, disk utilization, and central processing unit utilization.

show abstract

“…Optimizing caching policies involves answering the questions of what tiles to cache and where. [36] proposes a multi-neural network solution to maximize the cache hit ratio (CHR) of tiled 360° video transmission. Specifically, an LSTM network is used for video popularity predictions and a CNN network is adopted to perform content-based tile classification.…”

Section: Dl-based 360° Video Transmissionmentioning

confidence: 99%

A Review of Deep Learning Solutions in 360° Video Streaming

Mahmoud

Rizou

Panayides³

et al. 2023

2023 12th International Conference on Modern Circuits and Systems Technologies (MOCAST)

View full text Add to dashboard Cite

The spread of virtual reality and 360° video applications has raised research interest in developing new streaming techniques. On one hand, 360° videos rely on strict network requirements compared to conventional 2D videos. Realizing an adequate user experience is subject to ultra-low latency and huge bitrate requirements. On the other hand, 360° videos have distinct characteristics that allow for innovative streaming solutions. These solutions have benefited from the advancements in deep learning for optimizing the transmission under restricted network resources. In this paper, we review existing works employing deep learning in 360° video transmission and we highlight the challenges associated with 360° video streaming.

show abstract

Multi-neural network based tiled 360°video caching with Mobile Edge Computing

Cited by 16 publications

References 20 publications

Optimal Video Caching at The Edge of Network by Using Machine Learning

Optimal Video Caching at The Edge of Network by Using Machine Learning

OJUMP: Optimization for joint unicast‐multicast panoramic VR live streaming system

A Review of Deep Learning Solutions in 360° Video Streaming

Contact Info

Product

Resources

About