PRACTICAL USE CASES FOR CREATING CONVOLUTIONAL NEURAL NETWORK MODELS FOR IMAGE RECOGNITION TASKS
DOI:
https://doi.org/10.26906/SUNZ.2024.3.136Keywords:
computer vision, classification tasks, deep learning, neural networksAbstract
The aim of the article is to improve the efficiency of image recognition by creating a binary classifier for objects of civil infrastructure using deep learning techniques. Research results. Convolutional neural network models have been created for the recognition of civil infrastructure objects. A Sequential convolutional neural network has been constructed, consisting of three convolutional layers, pooling layers, a transformation layer, a fully connected layer, and an output layer. The optimal error values during training/testing are 0.0650/0.4424, with accuracies of 0.98/0.92 respectively. Results from the third epoch show errors of 0.2442/0.2595 and an accuracy of 0.93. A pre-trained VGG16 model was also utilized, fine-tuned on the dataset, demonstrating minimal error values of 0.0278/0.1538 during training/testing, with accuracies of 0.99/0.96 respectively. Scientific novelty. Further development of using convolutional neural networks for recognizing the level of civil infrastructure damage has been achieved. Practical significance. Two convolutional neural network models, Sequential and VGG16, have been built to address the recognition of damaged and intact buildings. The prerequisite for detecting these objects is the use of a camera and appropriate hardware such as a Raspberry Pi or a personal computer/laptop.Downloads
References
S V. Satellite Image Classification using CNN with Particle Swarm Optimization Classifier [Electronic resource] / Vidhya S, Balaji M, Kamaraj V // Procedia Computer Science. – 2024. – Vol. 233. – P. 979–987. – Mode of access: https://doi.org/10.1016/j.procs.2024.03.287
O'Shaughnessy D. Understanding Automatic Speech Recognition [Electronic resource] / Douglas O'Shaughnessy // Computer Speech & Language. – 2023. – P. 101538. – Mode of access: https://doi.org/10.1016/j.csl.2023.101538
Practical principles of integrating artificial intelligence into the technology of regional security predicting [Electronic resource] / Oleksandr Shefer [et al.] // Advanced Information Systems. – 2024. – Vol. 8, no. 1. – P. 86–93. – Mode of access: https://doi.org/10.20998/2522-9052.2024.1.11
Computing the characteristics of defects in wooden structures using image processing and CNN [Electronic resource] / Rana Ehtisham [et al.] // Automation in Construction. – 2024. – Vol. 158. – P. 105211. – Mode of access: https://doi.org/10.1016/j.autcon.2023.105211
Hybrid framework for respiratory lung diseases detection based on classical CNN and quantum classifiers from chest X-rays [Electronic resource] / G. V. Eswara Rao [et al.] // Biomedical Signal Processing and Control. – 2024. – Vol. 88. – P. 105567. – Mode of access: https://doi.org/10.1016/j.bspc.2023.105567
Artificial-intelligence-led revolution of construction materials: From molecules to Industry 4.0 [Electronic resource] / Xing Quan Wang [et al.] // Matter. – 2023. – Vol. 6, no. 6. – P. 1831–1859. – Mode of access: https://doi.org/10.1016/j.matt.2023.04.016
Yasmine G. Anti-drone systems: An Attention Based Improved YOLOv7 model for a real-time detection and identification of multi-airborne target [Electronic resource] / Ghazlane Yasmine, Gmira Maha, Medromi Hicham // Intelligent Systems with Applications. – 2023. – P. 200296. – Mode of access: https://doi.org/10.1016/j.iswa.2023.200296
Deep hypergraph autoencoder embedding: An efficient intelligent approach for rotating machinery fault diagnosis [Electronic resource] / Mingkuan Shi [et al.] // Knowledge-Based Systems. – 2022. – P. 110172. – Mode of access: https://doi.org/10.1016/j.knosys.2022.110172
Onyshchenko S. Improving the efficiency of diagnosing errors in computer devices for processing economic data functioning in the class of residuals [Electronic resource] / Svitlana Onyshchenko, Alina Yanko, Alina Hlushko // Eastern-European access: https://doi.org/10.15587/1729-4061.2023.289185
Lv X.-L. Visual clustering network-based intelligent power lines inspection system [Electronic resource] / Xian-Long Lv, Hsiao-Dong Chiang // Engineering Applications of Artificial Intelligence. – 2024. – Vol. 129. – P. 107572. – Mode of access: https://doi.org/10.1016/j.engappai.2023.107572
A hybrid CNN-GRU based probabilistic model for load forecasting from individual household to commercial building [Electronic resource] / Ming-Chuan Chiu [et al.] // Energy Reports. – 2023. – Vol. 9. – P. 94–105. – Mode of access: https://doi.org/10.1016/j.egyr.2023.05.090
Development of Shared Modeling and Simulation Environment for Sustainable e-Learning in the STEM Field [Electronic https://doi.org/10.3390/su16052197
Sikakollu P. Ensemble of multiple CNN classifiers for HSI classification with Superpixel Smoothing [Electronic resource] / Prasanth Sikakollu, Ratnakar Dash // Computers & Geosciences. – 2021. – Vol. 154. – P. 104806. – Mode of access:https://doi.org/10.1016/j.cageo.2021.104806
Cross-evaluation of a parallel operating SVM – CNN classifier for reliable internal decision-making processes in composite inspection [Electronic resource] / Sebastian Meister [et al.] // Journal of Manufacturing Systems. – 2021. – Vol. 60. – P. 620–639. – Mode of access: https://doi.org/10.1016/j.jmsy.2021.07.022
Hybridizing genetic random forest and self-attention based CNN-LSTM algorithms for landslide susceptibility mapping in Darjiling and Kurseong, India [Electronic resource] / Armin Moghimi [et al.] // Quaternary Science Advances. – 2024. – P. 100187. – Mode of access: https://doi.org/10.1016/j.qsa.2024.100187
Turkiye_Earthquake_2023 [Electronic resource] // Kaggle: Your Machine Learning and Data Science Community. – Mode of access: https://www.kaggle.com/datasets/buraktaci/turkiye-earthquake-2023?resource=download
François Chollet. Deep Learning with Python / François Chollet. – Shelter Island, NY : Manning Publications Co., 2018. – 361 p.
Introduction to the Keras Tuner | TensorFlow Core [Electronic resource] // TensorFlow. – Mode of access: https://www.tensorflow.org/tutorials/keras/keras_tuner
Netron [Electronic resource] // Netron. – Mode of access: https://netron.app/
Home [Electronic resource] // OpenCV. – Mode of access: https://opencv.org/
