AI Benchmark: All About Deep Learning on Smartphones in 2019

Ignatov, Andrey; Timofte, Radu; Kulik, Andrei; Yang, Seungsoo; Wang, Ke; Baum, Felix; Wu, Max; Xu, Lirong; Gool, Luc Van

doi:10.1109/iccvw.2019.00447

Cited by 180 publications

(101 citation statements)

References 52 publications

Supporting

Mentioning

101

Contrasting

Order By: Relevance

“…CNN-LSTM based models also has a larger decrease (≈ 3%) comparing to the LSTM. This is because the quantised inference was used in this work instead of the floating-point inference [29]. Quantised inference needs to convert the model from a 16-bit floating point type to int-8 format to reduce the size and random-access memory (RAM) consumption by a factor of 4 [29].…”

Section: ) Performance Of Selected Dnn Models With and Without Rule-based Reconstruction Methodsmentioning

confidence: 99%

“…This is why most of previous studies on resource-constraints systems, such as embedded device and smartphones, are only using CNNs. Recent advances in smartphone hardware and DNNs optimisation technologies on mobile platforms have significantly overcome this obstacle [29]. With increasing computing power on mobile devices, computationally intensive models such as LSTM can be run on a mobile device like a mobile phone.…”

Section: Related Workmentioning

confidence: 99%

“…TFLite library is part of Google's Tensorflow [49]. It offers a significantly reduced binary size and kernels optimized for on-device inference [29]. TFLite is also supported by Android Neural Networks API (NNAPI) which can utilise the on-device AI hardware acceleration resources (such as the mobile GPUs) directly to improve the inference performance.…”

Section: B Lpm Detection and Classification On Mobile Platformmentioning

confidence: 99%

“…However, TFLite is still lacking the support of some vital deep learning operators, especially for RNNs such as LSTM [50]. Therefore, this work used the TensorFlow Lite plugin, which contained the standard Tensorflow operators [29], to implement and convert these selected DNN models for LPM detection and classification on the mobile devices.…”

Section: B Lpm Detection and Classification On Mobile Platformmentioning

confidence: 99%

See 3 more Smart Citations

On-Device Lumbar-Pelvic Movement Detection Using Dual-IMU: A DNN-Based Approach

et al. 2021

View full text Add to dashboard Cite

Lumbar-pelvic movements (LPMs) are generally performed in the clinical setting to identify limitations in a range of movements. Continuous monitoring of these movements can provide real-time feedback to both patients and medical experts with the potential of identifying activities that may precipitate symptoms of low back pain (LBP) as well as improving therapy by providing a personalised approach. Recent advances in mobile computing technology and wearable sensors have paved the way for developing mobile physical activity monitoring applications with more advanced and complex algorithms, such as deep neural network (DNN) based models. However, there is a lack of prior studies that focus on real-time LPMs detection with multimodal sensory data. Meanwhile, most research in the area of body movement detection do not consider the potential transition logic of the constituent low-level body movements (e.g., LPMs) within their corresponding high-level physical activity. This information could significantly increase accuracy of detection results. Further, current studies mainly perform deep learning-based movement detection in the cloud (or a backend server) that could increase network bandwidth and response time.To address these limitations, this paper proposes a novel LPMs detection approach using an enhanced and adapted hybrid DNN model, which includes a convolutional neural network followed by a long-short term memory recurrent network (CNN-LSTM), and performs detection locally on the mobile device. The results of a comparative evaluation of the proposed model and baseline models are described. We also introduce a set of domain-specific pre-defined rules, based on the transition logic information, to reconstruct the detection outputs to further improve the detection accuracy.

show abstract

Section: ) Performance Of Selected Dnn Models With and Without Rule-based Reconstruction Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: B Lpm Detection and Classification On Mobile Platformmentioning

confidence: 99%

Section: B Lpm Detection and Classification On Mobile Platformmentioning

confidence: 99%

See 2 more Smart Citations

On-Device Lumbar-Pelvic Movement Detection Using Dual-IMU: A DNN-Based Approach

et al. 2021

View full text Add to dashboard Cite

show abstract

“…3.1.3 On-demand Image Delivery. As chipsets on commodity devices are gradually getting more powerful [2,20], this enables many applications to run fully on-device, avoiding the latency and privacy issues of offloading. In this direction, Lee et al [29] proposed MobiSR (Fig.…”

Section: °Videomentioning

confidence: 99%

Neural Enhancement in Content Delivery Systems

Lee

Venieris

Lane

2020

Proceedings of the 1st Workshop on Distributed Machine Learning

View full text Add to dashboard Cite

Internet-enabled smartphones and ultra-wide displays are transforming a variety of visual apps spanning from ondemand movies and 360°videos to video-conferencing and live streaming. However, robustly delivering visual content under fluctuating networking conditions on devices of diverse capabilities remains an open problem. In recent years, advances in the field of deep learning on tasks such as superresolution and image enhancement have led to unprecedented performance in generating high-quality images from low-quality ones, a process we refer to as neural enhancement. In this paper, we survey state-of-the-art content delivery systems that employ neural enhancement as a key component in achieving both fast response time and high visual quality. We first present the deployment challenges of neural enhancement models. We then cover systems targeting diverse use-cases and analyze their design decisions in overcoming technical challenges. Moreover, we present promising directions based on the latest insights from deep learning research to further boost the quality of experience of these systems. CCS CONCEPTS • Computing methodologies → Computer vision tasks; Distributed computing methodologies.

show abstract

Nodes selection review for federated learning in the blockchain‐based internet of things

Abdmeziem,

Akli,

Zourane

2023

Security and Privacy

View full text Add to dashboard Cite

Internet of Things (IoT) gained momentum these last few years pushed by the emergence of fast and reliable communication networks such as 5G and beyond. IoT depends on collecting information from the environment, leading to a significant increase in the amount of data generated that needs to be transmitted, saved, and analyzed. It is clear that classical deterministic approaches might not be suitable to this complex and fast evolving environment. Hence, machine learning techniques with their ability to handle such a dynamic context, are rising in popularity. In particular, Federated Learning architectures which are better suited to the distributed nature of IoT and its privacy concerns. Besides, to address security risks such as model poisoning, device compromise, and network interception, Blockchain (BC) is seen as the secure and distributed underlying communication infrastructure of choice. This integration of IoT, FL, and BC remains in its early stages and several challenges arise. Indeed, nodes selection to perform resource intensive and critical operations like model learning and transactions validation is a crucial issue considering the strong heterogeneity of the involved devices in terms of resources. In this paper, we propose an original literature review including a taxonomy, a thorough analysis, a comparison of the proposed approaches, along with some open research directions.

show abstract

AI Benchmark: All About Deep Learning on Smartphones in 2019

Cited by 180 publications

References 52 publications

On-Device Lumbar-Pelvic Movement Detection Using Dual-IMU: A DNN-Based Approach

On-Device Lumbar-Pelvic Movement Detection Using Dual-IMU: A DNN-Based Approach

Neural Enhancement in Content Delivery Systems

Nodes selection review for federated learning in the blockchain‐based internet of things

Contact Info

Product

Resources

About