Convolutional neural network hyperparameter optimization applied to land cover classification

Yaloveha, Vladyslav; Podorozhniak, Andrii; Kuchuk, Heorhii

doi:10.32620/reks.2022.1.09

Cited by 21 publications

(22 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Classified 3 emotions using both audio and text data from the IEMOCAP dataset. A modern optimization framework today is Bayesian optimization, which is used worldwide for those black-box functions that are expensive [15]. It tunes artificial neural network models to tasks such as speech recognition, image classification and language modeling [15].…”

Section: Related Workmentioning

confidence: 99%

“…A modern optimization framework today is Bayesian optimization, which is used worldwide for those black-box functions that are expensive [15]. It tunes artificial neural network models to tasks such as speech recognition, image classification and language modeling [15]. Two parts of Bayesian optimization: the surrogate model is used to model the objective function and the acquisition function that outputs a value that is determined by the objective function at a new point [15].…”

Section: Related Workmentioning

confidence: 99%

“…It tunes artificial neural network models to tasks such as speech recognition, image classification and language modeling [15]. Two parts of Bayesian optimization: the surrogate model is used to model the objective function and the acquisition function that outputs a value that is determined by the objective function at a new point [15]. Bayesian is a probabilistic model [15].…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Emotion recognition of human speech using deep learning method and MFCC features

Hazra

Ema²,

Galib³

et al. 2022

reks

View full text Add to dashboard Cite

Subject matter: Speech emotion recognition (SER) is an ongoing interesting research topic. Its purpose is to establish interactions between humans and computers through speech and emotion. To recognize speech emotions, five deep learning models: Convolution Neural Network, Long-Short Term Memory, Artificial Neural Network, Multi-Layer Perceptron, Merged CNN, and LSTM Network (CNN-LSTM) are used in this paper. The Toronto Emotional Speech Set (TESS), Surrey Audio-Visual Expressed Emotion (SAVEE) and Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) datasets were used for this system. They were trained by merging 3 ways TESS+SAVEE, TESS+RAVDESS, and TESS+SAVEE+RAVDESS. These datasets are numerous audios spoken by both male and female speakers of the English language. This paper classifies seven emotions (sadness, happiness, anger, fear, disgust, neutral, and surprise) that is a challenge to identify seven emotions for both male and female data. Whereas most have worked with male-only or female-only speech and both male-female datasets have found low accuracy in emotion detection tasks. Features need to be extracted by a feature extraction technique to train a deep-learning model on audio data. Mel Frequency Cepstral Coefficients (MFCCs) extract all the necessary features from the audio data for speech emotion classification. After training five models with three datasets, the best accuracy of 84.35 % is achieved by CNN-LSTM with the TESS+SAVEE dataset.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Emotion recognition of human speech using deep learning method and MFCC features

Hazra

Ema²,

Galib³

et al. 2022

reks

View full text Add to dashboard Cite

show abstract

“…Зазначимо, що моделі таких задач безпосередньо пов'язані із задачами покриття області (простору, територій, тощо). Мова йдеться про спостереження над космічним простором [4][5][6], моніторинг навколишнього середовища, лісових пожеж [7,8], планування аварійно-рятувальних робіт [9], в робототехніці [10] та промисловій діагностиці [11,12] тощо.…”

Section: с в яковлевunclassified

Формалізація та розв'язування задачі максимального покриття області з використанням бібліотеки Shapely для моніторингу територій

Yakovlev

Kartashov

Mumrienko

2022

reks

View full text Add to dashboard Cite

To ensure the life of society, it becomes necessary to create systems for monitoring processes or objects using a network of sensors that can control part of the space (territory). Monitoring is understood as a systematic observation of the parameters of an object to obtain information on their compliance with the initial assumptions. Simultaneously, a physical model is constructed that links the characteristics of the object and information about the observation, which making it possible to identify the properties of the object. Such information is based on the processing of signals received using special control sensors. These signals are digitized to provide data on the coverage areas of the sensors. Thus, the physical model is associated with measuring the ability and quality of perception of control sensors, fixing the geometric relationship between them and points in space. The specified physical model corresponds to the geometric statement of the problem of covering the monitoring area with a set of geometric objects, the shape and size of which is determined by the coverage areas of the sensors. With a limited number of sensors, the problem arises of the maximum possible coverage of the area. In this article, we digress from the type of monitoring object and consider the geometric features of coverage problems that arise when designing systems for monitoring a space of various purposes. The current article presents constructive means of mathematical modeling for solving geometric problems of maximum coverage. To formalize the coverage conditions, the concept of constructing the configuration space of geometric objects and a special class of functions are used to establish the dependence of the measure (area, volume) of the coverage configuration on the placement parameters of the covering objects. Since it is extremely difficult to obtain an analytical form of these functions, an algorithmic approach to their calculation is proposed. The approach was implemented on the Pyton algorithm using the Shapely library. A computational experiment was planned and carried out to establish the dependence of the computation time on the number of geometric objects that make up the coverage configuration. To find the maximum coverage, the BFGS local optimization method of the Scipy.optimize package is used. Numerous examples of the implemenation of the proposed approach are given. Conclusions. The article substantiates the use of a software-algorithmic approach for formalization, calculation and optimization of maximum coverage configurations, which makes it possible to effectively solve complex problems of monitoring space and territories.

show abstract

“…Analysis of latest researches and publications. Analysis of latest researches and publications shows that scientists all over the world make their researches in the field of image analysis to do practical tasks not only in the sphere of information technology, but in the sphere of forestry and agriculture [1][2][3][4][5][6].…”

Section: Introductionmentioning

confidence: 99%

Method of Identification of Tree Species Composition of Forests on the Basis of Geographic Information Database

Barabash

Bandurka

Svynchuk

et al. 2022

A.I.S.

View full text Add to dashboard Cite

Monitoring of the forests with the help of remote sensing of the Earth systems is actual on the present stage of people development. Existing systems of research do not allow to use geographic information databases and don't include the spectrum of exploration that are presented in our work. The aim of this article is to develop the method of identification of tree species composition of forests on the basis of geographic information database in the integration with correlation analysis. It will give the possibility to find the most negative things that influence on the condition of forest ecosystems and neutralize them at the first stage of their influence. Satellite images that are received from the satellites Landsat 5 and 8 are used to make the database. Tree species composition of Mozhariv forest during 20 years is analyzed as an example in this article. Classification of tree species composition of forests using space images is made on the basis of Bayes classifier. There is a special geographic information database for further data processing and identification of tree species composition is developed. Forest is analyzed on the number of trees in each species during continuous time measures (in 2000, 2010 and 2020) in this article. Results of the work are presented with the help of graphics. Correlation analysis, which helps to make the analysis of tightness of connections between changes trees in number and different forest criteria, is made to identify tree species composition of forest. Developed method together with well-known methods of monitoring of forests help us to make effective application for analysis of tree species composition of forest changes and to make better existing information systems including different quality demands. Peculiarity of developed information system is in using of the data processing module as a part of geographic information system and implementing correlation analysis with identification of factors that influence each other. It helps to increase quality in management and monitoring of forest resources. Geographic information becomes available for users.

show abstract

Convolutional neural network hyperparameter optimization applied to land cover classification

Cited by 21 publications

References 5 publications

Emotion recognition of human speech using deep learning method and MFCC features

Emotion recognition of human speech using deep learning method and MFCC features

Формалізація та розв'язування задачі максимального покриття області з використанням бібліотеки Shapely для моніторингу територій

Method of Identification of Tree Species Composition of Forests on the Basis of Geographic Information Database

Contact Info

Product

Resources

About