Mutual information guided 3D ResNet for self‐supervised video representation learning

Xue, Fei; Ji, Hongbing; Zhang, Wenbo

doi:10.1049/iet-ipr.2020.0019

Cited by 4 publications

(5 citation statements)

References 36 publications

(82 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Xue et al introduced a unique self-supervised learning technique based on ResNet, to enable automatic learning of video representations. Their suggested self-supervised learning approach was found to be effective and to have some degree of generalizability when utilized as an efficient pre-training technique for the job of identifying the activities in the video [8]. Salama et al suggested a ResNet-based breast cancer medical image-assisted diagnosis model and confirmed the model's efficacy.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Real-Time Management of Coal Mine Underground Shield Machine Digging Speed Based on Improved Residual Neural Networks

Xu,

Qi,

Liang

2024

IEEE Access

View full text Add to dashboard Cite

Aiming at the lack of accuracy and effectiveness of the current shield machine speed prediction method, the study proposes to improve the residual network and combine this improved algorithm with the surrounding rock category prediction model to construct the underground shield machine digging speed prediction model. With an average accuracy of 87.4%, an F1 value of 0.86, and an accuracy of 0.84, the study's prediction model of surrounding rock categories was determined to be valid and superior to the other compared models. The effectiveness of the improved residual algorithm constructed by the study was verified, and it was found to have a better fit to the actual values, with a maximum deviation error value of 4.6 mm/min and a root mean square error of 1.835, which was lower than the other comparative algorithms. The empirical analysis of the underground shield machine digging speed prediction model constructed by the study revealed that the area under the line of the work characteristic curve of the subjects was 0.74, and the F1 value was 0.35, and the accuracy was as high as 84.6%, which was significantly better than that of other comparative models. The shield machine digging speed prediction model, which is based on an enhanced residual network built in the study, performs better than other comparison models, according to the results, which can serve as a theoretical guide for the digital management of coal mine output.

show abstract

Section: Related Workmentioning

confidence: 99%

“…In equation (8), act denotes the activation function and θ denotes the weight data. b denotes the bias and '…”

Section: B Modified Resnet Algorithm Incorporating Lstmsmentioning

confidence: 99%

Real-Time Management of Coal Mine Underground Shield Machine Digging Speed Based on Improved Residual Neural Networks

Xu,

Qi,

Liang

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Our 3D ResNet 50 model employs 3D convolutions and 3D batch normalization. This spatiotemporal-feature-learning method enables the 3D ResNet 50 model to better capture the dynamic information in videos compared to 2D CNN, resulting in an enhanced ability to identify and differentiate the categories of videos 41,42 . Furthermore, we have adjusted the structure of identity shortcuts to reduce information loss during downsampling and reduced the frequency at which the model doubles the number of channels after passing through the residual structure (see Supplementary Information for the detailed structure of the ResNet 50 model and the principle of the basic residual block in ResNet, S5 and S6).…”

Section: Model Construction and Evaluationmentioning

confidence: 99%

Machine-learning-assisted and real-time-feedback-controlled growth of InAs/GaAs quantum dots

Shen,

Zhan,

Xin

et al. 2024

Nat Commun

View full text Add to dashboard Cite

The applications of self-assembled InAs/GaAs quantum dots (QDs) for lasers and single photon sources strongly rely on their density and quality. Establishing the process parameters in molecular beam epitaxy (MBE) for a specific density of QDs is a multidimensional optimization challenge, usually addressed through time-consuming and iterative trial-and-error. Here, we report a real-time feedback control method to realize the growth of QDs with arbitrary density, which is fully automated and intelligent. We develop a machine learning (ML) model named 3D ResNet 50 trained using reflection high-energy electron diffraction (RHEED) videos as input instead of static images and providing real-time feedback on surface morphologies for process control. As a result, we demonstrate that ML from previous growth could predict the post-growth density of QDs, by successfully tuning the QD densities in near-real time from 1.5 × 1010 cm−2 down to 3.8 × 108 cm−2 or up to 1.4 × 1011 cm−2. Compared to traditional methods, our approach can dramatically expedite the optimization process and improve the reproducibility of MBE. The concepts and methodologies proved feasible in this work are promising to be applied to a variety of material growth processes, which will revolutionize semiconductor manufacturing for optoelectronic and microelectronic industries.

show abstract

“…Based on the first-generation lightweight network mobile network vision 1 (MobileNetV1), the concepts of inverted residuals and linear bottlenecks are introduced into MobileNetV2 [27]. As a DW convolution cannot change the number of channels, feature extraction is restricted by the number of input channels.…”

Section: Mobilenetv2 Networkmentioning

confidence: 99%

Ancient mural segmentation based on a deep separable convolution network

et al. 2022

View full text Add to dashboard Cite

Traditional methods for ancient mural segmentation have drawbacks, including fuzzy target boundaries and low efficiency. Targeting these problems, this study proposes a pyramid scene parsing MobileNetV2 network (PSP-M) by fusing a deep separable convolution-based lightweight neural network with a multiscale image segmentation model. In this model, deep separable convolution-fused MobileNetV2, as the backbone network, is embedded in the image segmentation model, PSPNet. The pyramid scene parsing structure, particularly owned by the two models, is used to process the background features of images, which aims to reduce feature loss and to improve the efficiency of image feature extraction. In the meantime, atrous convolution is utilized to expand the perceptive field, aiming to ensure the integrity of image semantic information without changing the number of parameters. Compared with traditional image segmentation models, PSP-M increases the average training accuracy by 2%, increases the peak signal-to-noise ratio by 1–2 dB and improves the structural similarity index by 0.1–0.2.

show abstract

Mutual information guided 3D ResNet for self‐supervised video representation learning

Cited by 4 publications

References 36 publications

Real-Time Management of Coal Mine Underground Shield Machine Digging Speed Based on Improved Residual Neural Networks

Real-Time Management of Coal Mine Underground Shield Machine Digging Speed Based on Improved Residual Neural Networks

Machine-learning-assisted and real-time-feedback-controlled growth of InAs/GaAs quantum dots

Ancient mural segmentation based on a deep separable convolution network

Contact Info

Product

Resources

About