A Multi-View Human Action Recognition System in Limited Data Case using Multi-Stream CNN

Chenarlogh, Vahid Ashkani; Razzazi, Farbod; Mohammadyahya, Najmeh

doi:10.1109/icspis48872.2019.9066079

Cited by 25 publications

(10 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…No data from machine controls were used so far. Neither were more sophisticated approaches such as AI or Human Action Recognition applied yet [ 32 ]. This approach was chosen as in SMEs, often old and heterogeneous machine landscapes are found, and machine controls often do not offer data interfaces.…”

Section: Summary Conclusion and Further Researchmentioning

confidence: 99%

Enhanced Changeover Detection in Industry 4.0 Environments with Machine Learning

Miller,

Borysenko,

Heusinger

et al. 2021

Sensors

View full text Add to dashboard Cite

Changeover times are an important element when evaluating the Overall Equipment Effectiveness (OEE) of a production machine. The article presents a machine learning (ML) approach that is based on an external sensor setup to automatically detect changeovers in a shopfloor environment. The door statuses, coolant flow, power consumption, and operator indoor GPS data of a milling machine were used in the ML approach. As ML methods, Decision Trees, Support Vector Machines, (Balanced) Random Forest algorithms, and Neural Networks were chosen, and their performance was compared. The best results were achieved with the Random Forest ML model (97% F1 score, 99.72% AUC score). It was also carried out that model performance is optimal when only a binary classification of a changeover phase and a production phase is considered and less subphases of the changeover process are applied.

show abstract

Section: Summary Conclusion and Further Researchmentioning

confidence: 99%

Enhanced Changeover Detection in Industry 4.0 Environments with Machine Learning

Miller,

Borysenko,

Heusinger

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…As a subset of artificial intelligence, ML was widely used for tasks such as prediction, classification, and detection in various application scenarios [32][33][34][35][36][37][38]. Some of these applications are in fields of biomedical engineering [32,38], electrical engineering [33], petroleum engineering [34], computer engineering [35], urban planning engineering [36], and software engineering [37]. ML can also be useful in the context of telecommunication and LiFi.…”

Section: Application Of ML On Hybrid Ofdm Formsmentioning

confidence: 99%

A Novel OFDM Format and a Machine Learning Based Dimming Control for LiFi

2021

View full text Add to dashboard Cite

This paper proposes a new hybrid orthogonal frequency division multiplexing (OFDM) form termed as DC-biased pulse amplitude modulated optical OFDM (DPO-OFDM) by combining the ideas of the existing DC-biased optical OFDM (DCO-OFDM) and pulse amplitude modulated discrete multitone (PAM-DMT). The analysis indicates that the required DC-bias for DPO-OFDM-based light fidelity (LiFi) depends on the dimming level and the components of the DPO-OFDM. The bit error rate (BER) performance and dimming flexibility of the DPO-OFDM and existing OFDM schemes are evaluated using MATLAB tools. The results show that the proposed DPO-OFDM is power efficient and has a wide dimming range. Furthermore, a switching algorithm is introduced for LiFi, where the individual components of the hybrid OFDM are switched according to a target dimming level. Next, machine learning algorithms are used for the first time to find the appropriate proportions of the hybrid OFDM components. It is shown that polynomial regression of degree 4 can reliably predict the constellation size of the DCO-OFDM component of DPO-OFDM for a given constellation size of PAM-DMT. With the component switching and the machine learning algorithms, DPO-OFDM-based LiFi is power efficient at a wide dimming range.

show abstract

“…Due to training based on very large datasets, deep neural networks have a degree of accuracy vastly superior to that achieved using any other AI technologies in many areas, such as image recognition, natural language processing, text analysis, etc. [ 1 , 2 , 3 ]. Since accuracy is no longer an obstacle, deep neural network (DNN)-based AI technologies have become feasible for many applications.…”

Section: Introductionmentioning

confidence: 99%

An ASIP for Neural Network Inference on Embedded Devices with 99% PE Utilization and 100% Memory Hidden under Low Silicon Cost

Gao

Chen

Liu

2022

Sensors

View full text Add to dashboard Cite

The computation efficiency and flexibility of the accelerator hinder deep neural network (DNN) implementation in embedded applications. Although there are many publications on deep neural network (DNN) processors, there is still much room for deep optimization to further improve results. Multiple dimensions must be simultaneously considered when designing a DNN processor to reach the performance limit of the architecture, including architecture decision, flexibility, energy efficiency, and silicon cost minimization. Flexibility is defined as the ability to support as many multiple networks as possible and to easily adjust the scale. For energy efficiency, there are huge opportunities for power efficiency optimization, which involves access minimization and memory latency minimization based on on-chip memory minimization. Therefore, this work focused on low-power and low-latency data access with minimized silicon cost. This research was implemented based on an ASIP (application specific instruction set processor) in which an ISA was based on the caffe2 inference operator and the hardware design was based on a single instruction multiple data (SIMD) architecture. The scalability and system performance of our SoC extension scheme were demonstrated. The VLIW was used to execute multiple instructions in parallel. All costs for data access time were thus eliminated for the convolution layer. Finally, the processor was synthesized based on TSMC 65 nm technology with a 200 MHz clock, and the Soc extension scheme was analyzed in an experimental model. Our design was tested on several typical neural networks, achieving 196 GOPS at 200 MHz and 241 GOPS/W on the VGG16Net and AlexNet.

show abstract

A Multi-View Human Action Recognition System in Limited Data Case using Multi-Stream CNN

Cited by 25 publications

References 31 publications

Enhanced Changeover Detection in Industry 4.0 Environments with Machine Learning

Enhanced Changeover Detection in Industry 4.0 Environments with Machine Learning

A Novel OFDM Format and a Machine Learning Based Dimming Control for LiFi

An ASIP for Neural Network Inference on Embedded Devices with 99% PE Utilization and 100% Memory Hidden under Low Silicon Cost

Contact Info

Product

Resources

About