Research

Jump to: Review

21 pages, 1055 KiB

Open AccessArticle

Automatic Translation between Mixtec to Spanish Languages Using Neural Networks

by Hermilo Santiago-Benito , Diana-Margarita Córdova-Esparza , Noé-Alejandro Castro-Sánchez , Teresa García-Ramirez , Julio-Alejandro Romero-González and Juan Terven

Appl. Sci. 2024, 14(7), 2958; https://doi.org/10.3390/app14072958 - 31 Mar 2024

Viewed by 369

Abstract

This paper introduces a novel method for collecting and translating texts from the Mixtec to the Spanish language. The method comprises four primary steps. First, we collected a Mixtec–Spanish corpus that includes 4568 sentences from educational and religious domain texts. To enhance the [...] Read more.

This paper introduces a novel method for collecting and translating texts from the Mixtec to the Spanish language. The method comprises four primary steps. First, we collected a Mixtec–Spanish corpus that includes 4568 sentences from educational and religious domain texts. To enhance the parallel corpus, we generate synthetic data with GPT-3.5. Second, we cleaned the data with a semi-automatic approach followed by preprocessing and tokenization. In preprocessing, we removed stop words, duplicated sentences, special characters, and numbers and converted them to lowercase. Third, we performed semi-automatic alignment to find the correspondence of Mixtec–Spanish sentences to generate sentence-level aligned texts necessary for translation. Finally, we trained automatic translation models based on recurrent neural networks, bidirectional recurrent neural networks, and Transformers. Our system achieved a BLEU score of 95.66 for Mixtec-to-Spanish translation and 99.87 for Spanish-to-Mixtec translation. We also obtained a translation edit rate (TER) of 0.5 for Spanish-to-Mixtec and a TER of 16.5 for Mixtec-to-Spanish. Our research stands out as a pioneering effort in the field of automatic Mixtec-to-Spanish translation in Mexico, filling a gap identified in the current literature. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

23 pages, 10484 KiB

Open AccessArticle

Ball Tracking Based on Multiscale Feature Enhancement and Cooperative Trajectory Matching

by Xiao Han, Qi Wang and Yongbin Wang

Appl. Sci. 2024, 14(4), 1376; https://doi.org/10.3390/app14041376 - 07 Feb 2024

Viewed by 479

Abstract

Most existing object tracking research focuses on pedestrians and autonomous driving while ignoring sports scenes. When general object tracking models are used for ball tracking, there are often problems, such as detection omissions due to small object sizes and trajectory loss due to [...] Read more.

Most existing object tracking research focuses on pedestrians and autonomous driving while ignoring sports scenes. When general object tracking models are used for ball tracking, there are often problems, such as detection omissions due to small object sizes and trajectory loss due to occlusion. To address these challenges, we propose a ball detection and tracking model called HMMATrack based on multiscale feature enhancement and multilevel collaborative matching to improve ball-tracking results from the entire process of sampling, feature extraction, detection, and tracking. It includes a Heuristic Compound Sampling Strategy to deal with tiny sizes and imbalanced data samples; an MNet-based detection module to improve the ball detection accuracy; and a multilevel cooperative matching and automatic trajectory correction tracking algorithm that can quickly and accurately correct the ball’s trajectory. We also hand-annotated SportsTrack, a ball-tracking dataset containing soccer, basketball, and volleyball scenes. Extensive experiments are conducted on the SportsTrack, demonstrating that our proposed HMMATrack model outperforms other representative state-of-the-art models in ball detection and tracking. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

24 pages, 4130 KiB

Open AccessArticle

Random Forests Machine Learning Applied to PEER Structural Performance Experimental Columns Database

by Konstantinos G. Megalooikonomou and Grigorios N. Beligiannis

Appl. Sci. 2023, 13(23), 12821; https://doi.org/10.3390/app132312821 - 29 Nov 2023

Cited by 3 | Viewed by 1752

Abstract

Columns play a very important role in structural performance and, therefore, this paper contributes to the critical need for failure mode prediction of reinforced concrete (RC) columns by exploring the capabilities of random forest machine learning (ML) based on a well-known experimental column [...] Read more.

Columns play a very important role in structural performance and, therefore, this paper contributes to the critical need for failure mode prediction of reinforced concrete (RC) columns by exploring the capabilities of random forest machine learning (ML) based on a well-known experimental column database. Known as the PEER structural performance database, it assembles the results of over 400 cyclic, lateral-load tests of reinforced concrete columns. The database describes tests of spiral or circular hoop-confined columns, rectangular tied columns and columns with or without lap splices of longitudinal reinforcement at the critical sections. The efficiency towards the aforementioned goal of supervised ML methods such as random forests using a randomly assigned test set from the Pacific Earthquake Engineering Research Center (PEER) database is examined here. The overall accuracy score for rectangular RC columns is 94% and for circular RC columns is 86%. The latter performances are influenced by the size of the testing and training sets of data and are independent of the number of decision trees in the forest employed in the random forest algorithm. The performances of random forests in postdicting the failure mode of RC columns prove that ML has great promise in revolutionizing the profession of earthquake engineering. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

30 pages, 2053 KiB

Open AccessArticle

Determinants of Emotion Recognition System Adoption: Empirical Evidence from Malaysia

by Muhammad Nadzree Mohd Yamin, Kamarulzaman Ab. Aziz, Tan Gek Siang and Nor Azlina Ab. Aziz

Appl. Sci. 2023, 13(21), 11854; https://doi.org/10.3390/app132111854 - 30 Oct 2023

Viewed by 1693

Abstract

Emotion recognition systems (ERS) are an emerging technology with immense potential, exemplifying the innovative utilization of artificial intelligence (AI) within the context of the fourth industrial revolution (IR 4.0). Given that personalization is a key feature of the fifth industrial revolution (IR 5.0), [...] Read more.

Emotion recognition systems (ERS) are an emerging technology with immense potential, exemplifying the innovative utilization of artificial intelligence (AI) within the context of the fourth industrial revolution (IR 4.0). Given that personalization is a key feature of the fifth industrial revolution (IR 5.0), ERS has the potential to serve as an enabler for IR 5.0. Furthermore, the COVID-19 pandemic has increased the relevance of this technology as work processes were adapted for social distancing and the use of face masks. Even in the post-pandemic era, many individuals continue to wear face masks. Therefore, ERS offers a technological solution to address communication challenges in a masked world. The existing body of knowledge on ERS primarily focuses on exploring modalities or modes for emotion recognition, system development, and the creation of applications utilizing emotion recognition functions. However, to enhance the development of impactful ERS, it is essential for researchers and innovators to understand the factors that influence its usage and adoption among the intended users. Therefore, this study presents a framework that combines technology adoption theories to identify the determinants of ERS adoption among Malaysian youth. Data for this study were collected through a survey involving 386 respondents. The findings revealed attitudes, subjective norms, perceived behavioral control, and awareness as significant determinants of ERS adoption. Additionally, the study found that technology aptitude plays a moderating role. These insights can inform the formulation of effective policies and programs to encourage and facilitate the development of innovative ERS solutions. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

14 pages, 985 KiB

Open AccessArticle

Estimation of Obesity Levels with a Trained Neural Network Approach optimized by the Bayesian Technique

by Fatma Hilal Yagin, Mehmet Gülü, Yasin Gormez, Arkaitz Castañeda-Babarro, Cemil Colak, Gianpiero Greco, Francesco Fischetti and Stefania Cataldi

Appl. Sci. 2023, 13(6), 3875; https://doi.org/10.3390/app13063875 - 18 Mar 2023

Cited by 7 | Viewed by 2162

Abstract

Background: Obesity, which causes physical and mental problems, is a global health problem with serious consequences. The prevalence of obesity is increasing steadily, and therefore, new research is needed that examines the influencing factors of obesity and how to predict the occurrence [...] Read more.

Background: Obesity, which causes physical and mental problems, is a global health problem with serious consequences. The prevalence of obesity is increasing steadily, and therefore, new research is needed that examines the influencing factors of obesity and how to predict the occurrence of the condition according to these factors. This study aimed to predict the level of obesity based on physical activity and eating habits using the trained neural network model. Methods: The chi-square, F-Classify, and mutual information classification algorithms were used to identify the most critical factors associated with obesity. The models’ performances were compared using a trained neural network with different feature sets. The hyperparameters of the models were optimized using Bayesian optimization techniques, which are faster and more effective than traditional techniques. Results: The results predicted the level of obesity with average accuracies of 93.06%, 89.04%, 90.32%, and 86.52% for all features using the neural network and for the features selected by the chi-square, F-Classify, and mutual information classification algorithms. The results showed that physical activity, alcohol consumption, use of technological devices, frequent consumption of high-calorie meals, and frequency of vegetable consumption were the most important factors affecting obesity. Conclusions: The F-Classify score algorithm identified the most essential features for obesity level estimation. Furthermore, physical activity and eating habits were the most critical factors for obesity prediction. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

12 pages, 2241 KiB

Open AccessArticle

Robustness of Contrastive Learning on Multilingual Font Style Classification Using Various Contrastive Loss Functions

by Irfanullah Memon, Ammar ul Hassan Muhammad and Jaeyoung Choi

Appl. Sci. 2023, 13(6), 3635; https://doi.org/10.3390/app13063635 - 13 Mar 2023

Viewed by 1271

Abstract

Font is a crucial design aspect, however, classifying fonts is challenging compared with that of other natural objects, as fonts differ from images. This paper presents the application of contrastive learning in font style classification. We conducted various experiments to demonstrate the robustness [...] Read more.

Font is a crucial design aspect, however, classifying fonts is challenging compared with that of other natural objects, as fonts differ from images. This paper presents the application of contrastive learning in font style classification. We conducted various experiments to demonstrate the robustness of contrastive image representation learning. First, we built a multilingual synthetic dataset for Chinese, English, and Korean fonts. Next, we trained the model using various contrastive loss functions, i.e., normalized temperature scaled cross-entropy loss, triplet loss, and supervised contrastive loss. We made some explicit changes to the approach of applying contrastive learning in the domain of font style classification by not applying any image augmentation. We compared the results with those of a fully supervised approach and achieved comparable results using contrastive learning with fewer annotated images and a smaller number of training epochs. In addition, we also evaluated the effect of applying different contrastive loss functions on training. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

17 pages, 2972 KiB

Open AccessArticle

A Convolutional Autoencoder Approach for Boosting the Specificity of Retinal Blood Vessels Segmentation

by Natalia Nikoloulopoulou, Isidoros Perikos, Ioannis Daramouskas, Christos Makris, Povilas Treigys and Ioannis Hatzilygeroudis

Appl. Sci. 2023, 13(5), 3255; https://doi.org/10.3390/app13053255 - 03 Mar 2023

Cited by 5 | Viewed by 1212

Abstract

Automated retina vessel segmentation of the human eye plays a vital role as it can significantly assist ophthalmologists in identifying many eye diseases, such as diabetes, stroke, arteriosclerosis, cardiovascular disease, and many other human illnesses. The fast, automatic and accurate retina vessel segmentation [...] Read more.

Automated retina vessel segmentation of the human eye plays a vital role as it can significantly assist ophthalmologists in identifying many eye diseases, such as diabetes, stroke, arteriosclerosis, cardiovascular disease, and many other human illnesses. The fast, automatic and accurate retina vessel segmentation of the eyes is very desirable. This paper introduces a novel fully convolutional autoencoder for the retina vessel segmentation task. The proposed model consists of eight layers, each consisting of convolutional2D layers, MaxPooling layers, Batch Normalisation layers and more. Our model has been trained and evaluated on DRIVE and STARE datasets with 35 min of training time. The performance of the autoencoder model we introduce is assessed on two public datasets, the DRIVE and the STARE and achieved quite competitive results compared to the state-of-the-art methods in the literature. In particular, our model reached an accuracy of 95.73, an AUC_ROC of 97.49 on the DRIVE dataset, and an accuracy of 96.92 and an AUC ROC of 97.57 on the STARE dataset. Furthermore, our model has demonstrated the highest specificity among the methods in the literature, reporting a specificity of 98.57 on the DRIVE and 98.7 on the STARE dataset, respectively. The above statement can be noticed in the final blood vessel segmentation images produced by our convolutional autoencoder method since the segmentations are more accurate, sharp and noiseless than the result images of other proposed methods. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

27 pages, 2370 KiB

Open AccessArticle

Evaluating Deep Learning Techniques for Natural Language Inference

by Petros Eleftheriadis, Isidoros Perikos and Ioannis Hatzilygeroudis

Appl. Sci. 2023, 13(4), 2577; https://doi.org/10.3390/app13042577 - 16 Feb 2023

Viewed by 2157

Abstract

Natural language inference (NLI) is one of the most important natural language understanding (NLU) tasks. NLI expresses the ability to infer information during spoken or written communication. The NLI task concerns the determination of the entailment relation of a pair of sentences, called [...] Read more.

Natural language inference (NLI) is one of the most important natural language understanding (NLU) tasks. NLI expresses the ability to infer information during spoken or written communication. The NLI task concerns the determination of the entailment relation of a pair of sentences, called the premise and hypothesis. If the premise entails the hypothesis, the pair is labeled as an “entailment”. If the hypothesis contradicts the premise, the pair is labeled a “contradiction”, and if there is not enough information to infer a relationship, the pair is labeled as “neutral”. In this paper, we present experimentation results of using modern deep learning (DL) models, such as the pre-trained transformer BERT, as well as additional models that relay on LSTM networks, for the NLI task. We compare five DL models (and variations of them) on eight widely used NLI datasets. We trained and fine-tuned the hyperparameters for each model to achieve the best performance for each dataset, where we achieved some state-of-the-art results. Next, we examined the inference ability of the models on the BreakingNLI dataset, which evaluates the model’s ability to recognize lexical inferences. Finally, we tested the generalization power of our models across all the NLI datasets. The results of the study are quite interesting. In the first part of our experimentation, the results indicate the performance advantage of the pre-trained transformers BERT, RoBERTa, and ALBERT over other deep learning models. This became more evident when they were tested on the BreakingNLI dataset. We also see a pattern of improved performance when the larger models are used. However, ALBERT, given that it has 18 times fewer parameters, achieved quite remarkable performance. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

18 pages, 1890 KiB

Open AccessFeature PaperArticle

The Use of Digital Color Imaging and Machine Learning for the Evaluation of the Effects of Shade Drying and Open-Air Sun Drying on Mint Leaf Quality

by Ewa Ropelewska, Kadir Sabanci and Muhammet Fatih Aslan

Appl. Sci. 2023, 13(1), 206; https://doi.org/10.3390/app13010206 - 23 Dec 2022

Cited by 2 | Viewed by 1626

Abstract

The objective of this study was to reveal the usefulness of image processing and machine learning for the non-destructive evaluation of the changes in mint leaves caused by two natural drying techniques. The effects of shade drying and open-air sun drying on the [...] Read more.

The objective of this study was to reveal the usefulness of image processing and machine learning for the non-destructive evaluation of the changes in mint leaves caused by two natural drying techniques. The effects of shade drying and open-air sun drying on the ventral side (upper surface) and dorsal side (lower surface) of leaves were compared. Texture parameters were extracted from the digital color images converted to color channels R, G, B, L, a, b, X, Y, and Z. Models based on image features selected for individual color channels were built for distinguishing mint leaves in terms of drying techniques and leaf side using machine learning algorithms from groups of Lazy, Rules, and Trees. In the case of classification of the images of the ventral side of fresh and shade-dried mint leaves, an average accuracy of 100% and values of Precision, Recall, F-Measure, and MCC of 1.000 were obtained for color channels B (KStar and J48 machine learning algorithms), a (KStar and J48), b (KStar), and Y (KStar). The effect of open-air sun drying was greater. Images of the ventral side of fresh and open-air sun-dried mint leaves were completely correctly distinguished (100% correctness) for more color channels and algorithms, such as color channels R and G (J48), B, a and b (KStar, JRip, and J48), and X and Y (KStar). The classification of the images of the dorsal side of fresh and shade-dried mint leaves provided 100% accuracy in the case of color channel B (KStar) and a (KStar, JRip, and J48). The fresh and open-air sun-dried mint leaves imaged on the dorsal side were correctly classified at an accuracy of 100% for selected textures from color channels a (KStar, JRip, J48), b (J48), and Z (J48). The developed approach may be used in practice to monitor the changes in the structure of mint leaves caused by drying in a non-destructive, objective, cost-effective, and fast manner without the need to damage the leaves. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

17 pages, 1850 KiB

Open AccessArticle

A Variable Neighbourhood Search-Based Algorithm for the Transit Route Network Design Problem

by Christina Iliopoulou, Ioannis Tassopoulos and Grigorios Beligiannis

Appl. Sci. 2022, 12(20), 10232; https://doi.org/10.3390/app122010232 - 11 Oct 2022

Cited by 7 | Viewed by 1299

Abstract

The transit route network design problem (TRNDP) has long attracted research attention, with many metaheuristic approaches proposed for its solution. So far, and despite the promising performance of Variable Neighbourhood Search (VNS) variants for vehicle routing problems, the performance of the algorithm on [...] Read more.

The transit route network design problem (TRNDP) has long attracted research attention, with many metaheuristic approaches proposed for its solution. So far, and despite the promising performance of Variable Neighbourhood Search (VNS) variants for vehicle routing problems, the performance of the algorithm on the TRNDP remains unexplored. In this context, this study develops a VNS-based algorithm for the problem at hand. The performance of the algorithm is tested using benchmark networks used in bus transit network design and compared with some of the most recent and efficient methods from the literature. Results show that the algorithm yields superior results over existing implementations in short computational times. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

Review

Jump to: Research

26 pages, 1939 KiB

Open AccessReview

A Survey of Full-Cycle Cross-Modal Retrieval: From a Representation Learning Perspective

by Suping Wang, Ligu Zhu, Lei Shi, Hao Mo and Songfu Tan

Appl. Sci. 2023, 13(7), 4571; https://doi.org/10.3390/app13074571 - 04 Apr 2023

Cited by 2 | Viewed by 2466

Abstract

Cross-modal retrieval aims to elucidate information fusion, imitate human learning, and advance the field. Although previous reviews have primarily focused on binary and real-value coding methods, there is a scarcity of techniques grounded in deep representation learning. In this paper, we concentrated on [...] Read more.

Cross-modal retrieval aims to elucidate information fusion, imitate human learning, and advance the field. Although previous reviews have primarily focused on binary and real-value coding methods, there is a scarcity of techniques grounded in deep representation learning. In this paper, we concentrated on harmonizing cross-modal representation learning and the full-cycle modeling of high-level semantic associations between vision and language, diverging from traditional statistical methods. We systematically categorized and summarized the challenges and open issues in implementing current technologies and investigated the pipeline of cross-modal retrieval, including pre-processing, feature engineering, pre-training tasks, encoding, cross-modal interaction, decoding, model optimization, and a unified architecture. Furthermore, we propose benchmark datasets and evaluation metrics to assist researchers in keeping pace with cross-modal retrieval advancements. By incorporating recent innovative works, we offer a perspective on potential advancements in cross-modal retrieval. Full article

(This article belongs to the Special Issue Advances in Intelligent Information Systems and AI Applications)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Advances in Intelligent Information Systems and AI Applications

Share This Special Issue

Special Issue Editors

Special Issue Information

Keywords

Published Papers (11 papers)

Research

Review

Further Information

Guidelines

MDPI Initiatives

Follow MDPI