Necessary conditions for convergence of CNNs and initialization of convolution kernels

Hui-bin, Zhang; Feng, Liping; Zhang, Xiaohua; Yang, Yuchi; Li, Jing

doi:10.1016/j.dsp.2022.103397

Cited by 23 publications

(6 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Para [31] las redes neuronales artificiales "están formadas por capas de variables interconectadas entre sí con unos pesos que el algoritmo va calculando y ajustando mediante iteraciones (repeticiones) de un proceso que se conoce como «descenso del gradiente»". Para [32] atribuye el éxito de las redes neuronales artificiales en usar varios algoritmos de optimización y varias estructuras de la red, como la normalización por lotes.…”

Section: B Red Neuronal Artificialunclassified

Deep Learning to support the recognition of pests and diseases in Yungay potato crops in the province of Cutervo, Angurra hamlet, Perú

Huarote Zegarra,

Cabrera Herrera,

Llanos Chacaltana

2023

Proceedings of the 21th LACCEI International Multi-Conference for Engineering, Education and Technology (LACCEI 2023): “Leaders

View full text Add to dashboard Cite

The present investigation efficiently covers the need to classify according to the diseases or pests of the Yungay potato leaf in the province of Cutervo, Angurra village (Peru),specifically classify them as fly plague (PM=1), weevil plague (PG=2), streak disease (ER=3) or as no plague (SP=0). To achieve this, the digital images have necessarily been prepared to decrease in size and at the same time obtain their representative, this in order to enter the SOM neural network (self-organizing map). The functions based on artificial vision for the preparation of the image (in jpg format of varied dimensions), to the shots of the potato leaves, are 1200 images for training and another 1200 for tests, only the image being extracted from each scene. sheet, scaled to a dimension of 256x256 pixels to homogenize, extract its characteristics from each disease and passed to a tone of gray to be learned in the neural network, managing to verify with this sequence an accuracy of 99.42%, a

show abstract

Section: B Red Neuronal Artificialunclassified

Deep Learning to support the recognition of pests and diseases in Yungay potato crops in the province of Cutervo, Angurra hamlet, Perú

Huarote Zegarra,

Cabrera Herrera,

Llanos Chacaltana

2023

Proceedings of the 21th LACCEI International Multi-Conference for Engineering, Education and Technology (LACCEI 2023): “Leaders

View full text Add to dashboard Cite

show abstract

“…Convolutional neural networks (CNN) have been essential to the development of deep learning. Remarkable advancements have been made on image classification [34]. CNN architecture is mainly divided into convolution layer, pooling layer and fully connected layer (Figure 6).…”

Section: Deep Learning 241 Convolutional Neural Networkmentioning

confidence: 99%

Classification of Maize Lodging Extents Using Deep Learning Algorithms by UAV-Based RGB and Multispectral Images

et al. 2022

View full text Add to dashboard Cite

Lodging depresses the grain yield and quality of maize crop. Previous machine learning methods are used to classify crop lodging extents through visual interpretation and sensitive features extraction manually, which are cost-intensive, subjective and inefficient. The analysis on the accuracy of subdivision categories is insufficient for multi-grade crop lodging. In this study, a classification method of maize lodging extents was proposed based on deep learning algorithms and unmanned aerial vehicle (UAV) RGB and multispectral images. The characteristic variation of three lodging extents in RGB and multispectral images were analyzed. The VGG-16, Inception-V3 and ResNet-50 algorithms were trained and compared depending on classification accuracy and Kappa coefficient. The results showed that the more severe the lodging, the higher the intensity value and spectral reflectance of RGB and multispectral image. The reflectance variation in red edge band were more evident than that in visible band with different lodging extents. The classification performance using multispectral images was better than that of RGB images in various lodging extents. The test accuracies of three deep learning algorithms in non-lodging based on RGB images were high, i.e., over 90%, but the classification performance between moderate lodging and severe lodging needed to be improved. The test accuracy of ResNet-50 was 96.32% with Kappa coefficients of 0.9551 by using multispectral images, which was superior to VGG-16 and Inception-V3, and the accuracies of ResNet-50 on each lodging subdivision category all reached 96%. The ResNet-50 algorithm of deep learning combined with multispectral images can realize accurate lodging classification to promote post-stress field management and production assessment.

show abstract

“…Para [35] las redes neuronales artificiales "están formadas por capas de variables interconectadas entre sí con unos pesos que el algoritmo va calculando y ajustando mediante iteraciones (repeticiones) de un proceso que se conoce como «descenso del gradiente»". Para [36] atribuye el éxito de las redes neuronales artificiales en usar varios algoritmos de optimización y varias estructuras de la red, como la normalización por lotes.…”

Section: A Red Neuronal Artificialunclassified

Strategy for the detection of types of eye diseases using SOM neural network

Zegarra¹,

Lujan²,

Flores³

et al. 2022

Proceedings of the 20th LACCEI International Multi-Conference for Engineering, Education and Technology: “Education, Research A

View full text Add to dashboard Cite

This research aims to cover a need to be able to classify according to the funds of eyes in diabetic retinopathy disease, how to convert to gray tone, perform an equalization, apply the canny edge highlighting algorithm and apply morphological operations so that a SOM (self-organization map) neural network can be entered and classified. To achieve this, it is classified as 0 to diabetic retinopathy, 1 to glaucoma and 3 to healthy eyes. To corroborate this strategy, a public database of Fundus-images has been taken, being 45 images of eyes for training and for tests 15 images that were not part of the training were used and for the tests 3 images that were not part of the training were used and each grayscale image is scaled to a dimension of 256x256 pixels, managing to demonstrate with this strategy an affectivity of 93.7% certainty in the identification of class of eye disease.

show abstract

Necessary conditions for convergence of CNNs and initialization of convolution kernels

Cited by 23 publications

References 1 publication

Deep Learning to support the recognition of pests and diseases in Yungay potato crops in the province of Cutervo, Angurra hamlet, Perú

Deep Learning to support the recognition of pests and diseases in Yungay potato crops in the province of Cutervo, Angurra hamlet, Perú

Classification of Maize Lodging Extents Using Deep Learning Algorithms by UAV-Based RGB and Multispectral Images

Strategy for the detection of types of eye diseases using SOM neural network

Contact Info

Product

Resources

About