String representations and distances in deep Convolutional Neural Networks for image classification

Barat, Cécile; Ducottet, Christophe

doi:10.1016/j.patcog.2016.01.007

Cited by 55 publications

(12 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The transformation types include translation, reflection, and light and color transformation. In contrast to previous network models, AlexNet expands storage capacity by increasing the number of GPUs [22]. It uses at least two GPUs to save kernels on average, and the GPUs communicate in the specified layers.…”

Section: B Alexnet Convolutional Neural Networkmentioning

confidence: 99%

An Intelligent Image Feature Recognition Algorithm With Hierarchical Attribute Constraints Based on Weak Supervision and Label Correlation

et al. 2020

View full text Add to dashboard Cite

The ability to extract image features largely determines the accuracy of image classification. However, external interferences in images such as translation, rotation, scaling, occlusion, light, and nonlinear deformation, result in greater intra-class differences and inter-class similarity, which substantially increases the difficulty of image classification. Benefitting from the excellent ability of feature learning and extraction, Convolutional Neural Networks (CNNs) have achieved good results in the field of image classification. However, they are also characterized by a complex training process and high-demand parameter adjustment. This paper propose a convolutional neural network optimization method to optimize the model parameters with hierarchical attribute constraints and achieve intelligent recognition of specific image features. Based on the optimized model, we further construct weak supervision and label correlation optimization models and provide a visualization solution for feature recognition results. Experimental results demonstrated that the proposed algorithm can efficiently realize intelligent image feature recognition with high accuracy.

show abstract

Section: B Alexnet Convolutional Neural Networkmentioning

confidence: 99%

An Intelligent Image Feature Recognition Algorithm With Hierarchical Attribute Constraints Based on Weak Supervision and Label Correlation

et al. 2020

View full text Add to dashboard Cite

show abstract

“…For example, deep CNNs have been widely used for various computer vision tasks, such as large-scale detection and classification of object categories [2][3][4][5][6], deep CNN-based transfer learning [7,8] and speech recognition [9]. RNNs, another important branch of the deep neural networks family, were mainly designed for sequence modeling.…”

Section: Introductionmentioning

confidence: 99%

Convolutional Recurrent Neural Networks forHyperspectral Data Classification

2017

View full text Add to dashboard Cite

Deep neural networks, such as convolutional neural networks (CNN) and stacked autoencoders, have recently been successfully used to extract deep features for hyperspectral data classification. Recurrent neural networks (RNN) are another type of neural networks, which are widely used for sequence analysis because they are constructed to extract contextual information from sequences by modeling the dependencies between different time steps. In this paper, we study the ability of RNN for hyperspectral data classification by extracting the contextual information from the data. Specifically, hyperspectral data are treated as spectral sequences, and an RNN is used to model the dependencies between different spectral bands. In addition, we propose to use a convolutional recurrent neural network (CRNN) to learn more discriminative features for hyperspectral data classification. In CRNN, a few convolutional layers are first learned to extract middle-level and locally-invariant features from the input data, and the following recurrent layers are then employed to further extract spectrally-contextual information from the features generated by the convolutional layers. Experimental results on real hyperspectral datasets show that our method provides better classification performance compared to traditional methods and other state-of-the-art deep learning methods for hyperspectral data classification.

show abstract

Section: Introductionmentioning

confidence: 99%

“…With the continuous development of deep learning techniques, convolutional neural networks (CNNs) have demonstrated significant advantages in the field of object recognition, such as string recognition (Barat and Ducottet, 2016), script identification (Shi et al, 2016), 3D object retrieval (Leng et al, 2015), and human-action recognition (Xu et al, 2016b). First proposed by Lecun et al (LeCun and Bengio, 1995;LeCun, 1989), CNNs can be considered as a type of supervised highly nonlinear mapping that output targetfeatures in a specified format based on the input data.…”

Section: Introductionmentioning

confidence: 99%

Convolutional Neural Network for Asphalt Pavement Surface Texture Analysis

Zheng

Gao

Sha

et al. 2018

Computer aided Civil Eng

View full text Add to dashboard Cite

Several data processing techniques (DPTs) have been implemented for evaluating pavement surface texture to partially replace onsite inspections by humans. However, the extensively varying real‐world situations have resulted in challenges in the widespread adoption of DPTs. To overcome these challenges, we propose the use of a convolutional neural network (CNN) to calculate the mean texture depth (MTD) without computing the surface texture feature statistics. Because a CNN is capable of automatically learning data features, the proposed method does not require the conjugation of DPTs for extracting features. The proposed CNN was trained and tested using 8,000 and 1,000 3D scan data samples, respectively, and achieved an average error of 0.0024 cm. The stability of the CNN was analyzed based on various test results. Comparative studies were conducted to verify the superiority of the CNN over conventional MTD algorithms. The results demonstrated that the CNN‐based method is significantly more stable in various real‐world situations. Additionally, the CNN‐based method achieved higher accuracy of automatic feature extraction than traditional MTD methods. Finally, the CNN‐based method was applied to evaluate the surface texture statistics of four highways in Shanxi, China, which were different from the training and testing samples; the results establish the transferability of this method to different highways.

show abstract

String representations and distances in deep Convolutional Neural Networks for image classification

Cited by 55 publications

References 32 publications

An Intelligent Image Feature Recognition Algorithm With Hierarchical Attribute Constraints Based on Weak Supervision and Label Correlation

An Intelligent Image Feature Recognition Algorithm With Hierarchical Attribute Constraints Based on Weak Supervision and Label Correlation

Convolutional Recurrent Neural Networks forHyperspectral Data Classification

Convolutional Neural Network for Asphalt Pavement Surface Texture Analysis

Contact Info

Product

Resources

About