Hyperparameter tuning of convolutional neural networks for building construction image classification

被引:7
作者
Carvalho Ottoni, Andre Luiz [1 ]
Novo, Marcela Silva [2 ]
Costa, Dayana Bastos [3 ]
机构
[1] Univ Fed Reconcavo Bahia, Technol & Exact Ctr, Cruz Das Almas, Brazil
[2] Univ Fed Bahia, Dept Elect & Comp Engn, Salvador, BA, Brazil
[3] Univ Fed Bahia, Dept Struct & Construct Engn, Salvador, BA, Brazil
关键词
Deep learning; Convolutional neural networks; Hyperparameter tuning; Scott-Knott method; Building construction image classification; LEARNING APPROACH; DEEP; SYSTEM;
D O I
10.1007/s00371-021-02350-9
中图分类号
TP31 [计算机软件];
学科分类号
081202 ; 0835 ;
摘要
Deep Learning models have important applications in image processing. However, one of the challenges in this field is the definition of hyperparameters. Thus, the objective of this work is to propose a rigorous methodology for hyperparameter tuning of Convolutional Neural Network for building construction image classification, especially in roofs structure analysis. For this, the HyperTuningSK algorithm was developed, intended to create recommendation rankings for two hyperparameters: learning rate and optimizer. The approach uses concepts from the statistical design of experiments, such as Analysis of Variance and the Scott-Knott clustering algorithm. In addition, the adopted database includes images of inspections on buildings roofs made with unmanned aerial vehicles. The images are divided into two classes: (i) roofs with clean gutters and (ii) roofs with dirty gutters. The methods recommended by the HyperTuningSK algorithm achieved good results in comparison to the hyperparameters adopted in the literature. In this respect, adagrad015 achieved the highest average values of accuracy in the validation (100%) and testing steps (90%) for Convolutional Neural Network architecture with 12 layers. In addition, the hyperparameters recommended by the HyperTuningSK algorithm achieved the best test results for other two literature architectures: Densenet121 (85.7%) and VGG16 (84.4%).
引用
收藏
页码:847 / 861
页数:15
相关论文
共 79 条
[1]   Efficient object tracking using hierarchical convolutional features model and correlation filters [J].
Abbass, Mohammed Y. ;
Kwon, Ki-Chul ;
Kim, Nam ;
Abdelwahab, Safey A. ;
El-Samie, Fathi E. Abd ;
Khalaf, Ashraf A. M. .
VISUAL COMPUTER, 2021, 37 (04) :831-842
[2]   Using CNN for facial expression recognition: a study of the effects of kernel size and number of filters on accuracy [J].
Agrawal, Abhinav ;
Mittal, Namita .
VISUAL COMPUTER, 2020, 36 (02) :405-412
[3]   A meta-learning approach for selecting image segmentation algorithm [J].
Aguiar, Gabriel Jonas ;
Mantovani, Rafael Gomes ;
Mastelini, Saulo M. ;
de Carvalho, Andre C. P. F. L. ;
Campos, Gabriel F. C. ;
Barbon Junior, Sylvio .
PATTERN RECOGNITION LETTERS, 2019, 128 :480-487
[4]  
[Anonymous], 2018, Deep Learningwith R
[5]   Encoder-decoder network for pixel-level road crack detection in black-box images [J].
Bang, Seongdeok ;
Park, Somin ;
Kim, Hongjo ;
Kim, Hyoungkwan .
COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING, 2019, 34 (08) :713-727
[6]   Properties of sufficiency and statistical tests [J].
Bartlett, MS .
PROCEEDINGS OF THE ROYAL SOCIETY OF LONDON SERIES A-MATHEMATICAL AND PHYSICAL SCIENCES, 1937, 160 (A901) :0268-0282
[7]  
Basgalupp M.P., 2020, IN PRESS, P1
[8]   A survey on deep multimodal learning for computer vision: advances, trends, applications, and datasets [J].
Bayoudh, Khaled ;
Knani, Raja ;
Hamdaoui, Faycal ;
Mtibaa, Abdellatif .
VISUAL COMPUTER, 2022, 38 (08) :2939-2970
[9]   Face recognition in unconstrained environment with CNN [J].
Ben Fredj, Hana ;
Bouguezzi, Safa ;
Souani, Chokri .
VISUAL COMPUTER, 2021, 37 (02) :217-226
[10]   Evaluation of Deep Learning CNN Model for Land Use Land Cover Classification and Crop Identification Using Hyperspectral Remote Sensing Images [J].
Bhosle, Kavita ;
Musande, Vijaya .
JOURNAL OF THE INDIAN SOCIETY OF REMOTE SENSING, 2019, 47 (11) :1949-1958