About pyramid structure in convolutional neural networks

Ullah, Ihsan; Petrosino, Alfredo

doi:10.1109/ijcnn.2016.7727350

Cited by 22 publications

(16 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the CIFAR-10 dataset, 10000 training samples and 2000 test samples were used. As in [19], the optimized models were compared to some of the stateof-the-art models, namely, LENET [20] and the Caffe Cifar-10 model (C10). In addition, two pyramidal structured models presented in [19]-SPyr_Rev_LENET and SPyr_Rev_C10-were also evaluated.…”

Section: Resultsmentioning

confidence: 99%

“…As in [19], the optimized models were compared to some of the stateof-the-art models, namely, LENET [20] and the Caffe Cifar-10 model (C10). In addition, two pyramidal structured models presented in [19]-SPyr_Rev_LENET and SPyr_Rev_C10-were also evaluated. To provide fair comparisons, all models were trained from scratch using the same training parameters, for a maximum of 100 epochs early stopping when no improvement on the test accuracy was observed for 10 consecutive epochs.…”

Section: Resultsmentioning

confidence: 99%

“…Together with Pareto efficiency, this prediction prevents the waste of computational resources by not evaluating potentially bad solutions. In [19], another approach to minimize the number of hyperparameters of a deep learning model is presented. In their work, the authors demonstrate how pyramidal-structured convolutional neural networks present competitive results when compared to state-of-theart models, while scaling down the number of parameters and reducing the use of computational resources.…”

Section: Background and Related Workmentioning

confidence: 99%

See 2 more Smart Citations

An Experiment on the Use of Genetic Algorithms for Topology Selection in Deep Learning

Mattioli

Caetano

Cardoso

et al. 2019

Journal of Electrical and Computer Engineering

View full text Add to dashboard Cite

The choice of a good topology for a deep neural network is a complex task, essential for any deep learning project. This task normally demands knowledge from previous experience, as the higher amount of required computational resources makes trial and error approaches prohibitive. Evolutionary computation algorithms have shown success in many domains, by guiding the exploration of complex solution spaces in the direction of the best solutions, with minimal human intervention. In this sense, this work presents the use of genetic algorithms in deep neural networks topology selection. The evaluated algorithms were able to find competitive topologies while spending less computational resources when compared to state-of-the-art methods.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Background and Related Workmentioning

confidence: 99%

See 1 more Smart Citation

An Experiment on the Use of Genetic Algorithms for Topology Selection in Deep Learning

Mattioli

Caetano

Cardoso

et al. 2019

Journal of Electrical and Computer Engineering

View full text Add to dashboard Cite

show abstract

“…In essence, this allows the network to compress the lower level features in the early layers in order to make easy classification decisions in the later layers. Pyramid structures have been shown to produce better results …”

Section: Convolutional Neural Networkmentioning

confidence: 99%

Convolutional neural networks for automated damage recognition and damage type identification

Modarres

Astorga

Droguett

et al. 2018

Struct Control Health Monit

141

View full text Add to dashboard Cite

Recurring expenses associated with preventative maintenance and inspection produce operational inefficiencies and unnecessary spending. Human inspectors may submit inaccurate damage assessments and physically inaccessible locations, like underground mining structures, and pose additional logistical challenges. Automated systems and computer vision can significantly reduce these challenges and streamline preventative maintenance and inspection.The authors propose a convolutional neural network (CNN)-based approach to identify the presence and type of structural damage. CNN is a deep feed-forward artificial neural network that utilizes learnable convolutional filters to identify distinguishing patterns present in images. CNN is invariant to image scale, location, and noise, which makes it robust to classify damage of different sizes or shapes. The proposed approach is validated with synthetic data of a composite sandwich panel with debonding damage, and crack damage recognition is demonstrated on real concrete bridge crack images. CNN outperforms several other machine learning algorithms in completing the same task. The authors conclude that CNN is an effective tool for the detection and type identification of damage.

show abstract

“…For the CIFAR-10 dataset, 10000 training samples and 2000 test samples were used. As in [Ullah and Petrosino 2016], the optimized models were compared to some of the state-of-the-art models, namely LENET [LeCun et al 1998] and the Caffe Cifar-10 model (C10). In addition, two pyramidal structured models presented in [Ullah and Petrosino 2016] -…”

Section: Evolutionary Selection Of Dnn Topologiesmentioning

confidence: 99%

Deep visual tracking for fiducial markers: a pilot study in telerehabilitation

Mattioli¹

View full text Add to dashboard Cite

In the past few decades, technology evolution has brought many improvements to the field of human rehabilitation. Among these improvements, the delivery of rehabilitation services to users at distant locations-known as telerehabilitation-made it possible to provide patients with home-based therapy, reducing the need for displacement to rehabilitation centers. In addition, the use of virtual and augmented training environments have enabled realistic simulation in the rehabilitation process, enhancing patients' experience and reducing exposure to real-world risks. Despite these benefits, when used in a distributed context, Augmented Reality technology brings additional challenges to the development process, including the integration of software components provided by heterogeneous sources. Also, the development of robust tracking techniques, less prone to detection errors resulting from network transmission is desired. This particular requirement motivated the development of alternative tracking algorithms, including the use of deep neural networks in visual tracking. When designing these trackers, the major role of network's topology selection has lead to the investigation of automatic search techniques, including evolutionary computation algorithms. In this context, this work presents as contribution a software architecture model for Augmented Reality, focusing on the reliable integration of independent software components for augmented content streaming. In addition, deep visual trackers for fiducial marker detection were designed and evaluated, outperforming other trackers when processing remotely captured video frames. Finally, the application of an evolutionary algorithm in deep neural networks' topology selection is discussed, presenting competitive results when compared to state-of-the-art methods.

show abstract

About pyramid structure in convolutional neural networks

Cited by 22 publications

References 19 publications

An Experiment on the Use of Genetic Algorithms for Topology Selection in Deep Learning

An Experiment on the Use of Genetic Algorithms for Topology Selection in Deep Learning

Convolutional neural networks for automated damage recognition and damage type identification

Deep visual tracking for fiducial markers: a pilot study in telerehabilitation

Contact Info

Product

Resources

About