Iván Galo Reyes-Chacón
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Iván Galo Reyes-Chacón has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Manuel Eugenio Morocho-Cayamcela
Deep Learning for Autonomous Driving, Robotics, and Computer Vision Research Group (DeepARC Research), School of Mathematical and Computational Sciences, Yachay Tech University, Hda. San José s/n y Proyecto Yachay, Urcuquí 100119, Ecuador
13 shared publications - Paulina Vizcaíno-Imacaña
Faculty of Technical Sciences, School of Computer Science, UIDE-International University of Ecuador, Quito 170501, Ecuador
13 shared publications - Ariana Deyaneira Jiménez-Narváez
Yachay Tech University
3 shared publications - Juan Pablo Astudillo León
Communication Networks and Intelligent Services Research Group (ComNet Innova YT), School of Mathematical and Computational Sciences, Yachay Tech University, Hda. San José s/n y Proyecto Yachay, Urcuquí 100119, Ecuador
3 shared publications - Clayanela J. Zambrano-Caicedo
Deep Learning for Autonomous Driving, Robotics, and Computer Vision Research Group (DeepARC Research), School of Mathematical and Computational Sciences, Yachay Tech University, Urcuquí 100119, Ecuador
2 shared publications - Isidro R. Amaro
Yachay Tech University
2 shared publications
Latest Publications
- Open Access
Economic Prominence and Announced Additional Time in the 2026 FIFA World Cup: A Census-Based Observational Study
Referee discretion in football may respond to contextual and organizational pressures, yet whether similar asymmetries appear in minimum announced additional time remains unclear. We analyze all 104 matches of the 2026 FIFA World Cup to test whether the economic prominence differential between national teams is associated with the fourth official’s minimum board-announced second-half additional time (BoardAT). Economic prominence is measured using standardized log gross domestic product (GDP), and Poisson models adjust for observable match characteristics. When the economically more prominent team was losing at minute 80, the estimated slope was (clustered standard error (SE) , ; incidence-rate ratio (IRR) , 95% confidence interval (CI) ). The losing-versus-winning contrast was also not significant (Holm-adjusted ), and robustness checks produced similar results. These findings provide no statistically detectable evidence that national economic prominence was systematically associated with minimum announced second-half additional time in this tournament.Open AccessEconomic Prominence and Announced Additional Time in the 2026 FIFA World Cup: A Census-Based Observational Study
Referee discretion in football may respond to contextual and organizational pressures, yet whether similar asymmetries appear in minimum announced additional time remains unclear. We analyze all 104 matches of the 2026 FIFA World Cup to test whether the economic prominence differential between national teams is associated with the fourth official’s minimum board-announced second-half additional time (BoardAT). Economic prominence is measured using standardized log gross domestic product (GDP), and Poisson models adjust for observable match characteristics. When the economically more prominent team was losing at minute 80, the estimated slope was (clustered standard error (SE) , ; incidence-rate ratio (IRR) , 95% confidence interval (CI) ). The losing-versus-winning contrast was also not significant (Holm-adjusted ), and robustness checks produced similar results. These findings provide no statistically detectable evidence that national economic prominence was systematically associated with minimum announced second-half additional time in this tournament. - Open Access
- 16 October 2026
- Published by Atlantis Press SARL
- p. 257-266
Semi-Automatic Satelliten Image Segmentation for Urban Expansion Analysis in Santa Cruz Island, Galápagos
- 16 October 2026
- Published by Atlantis Press SARL
- p. 257-266
Open AccessSemi-Automatic Satelliten Image Segmentation for Urban Expansion Analysis in Santa Cruz Island, Galápagos
- 13 October 2026
- Published by Springer Nature
- p. 80-93
Pre-visual Detection of Latent Ecological Stress in Galapagos Flora Using YOLO-Seg and Sentinel-2
- 13 October 2026
- Published by Springer Nature
- p. 80-93
Pre-visual Detection of Latent Ecological Stress in Galapagos Flora Using YOLO-Seg and Sentinel-2
- Open Access
- 18 September 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 153 (1) , 9
Improving Pedestrian Age and Gender Classification in a DeepStream-Based Video Analytics Pipeline
Real-time pedestrian analytics systems often combine detection, tracking, and demographic classification. However, generic demographic classifiers may generalize poorly to deployment-specific camera conditions. This paper presents a DeepStream-based pipeline using PeopleNet, object tracking, and independent ResNet18 classifiers for gender and age estimation. A target-domain dataset of 3508 labeled pedestrian crops was used for fine-tuning. The gender model used validation-based threshold calibration, while age optimization included imbalance-aware training and external data balancing. On the target-domain test set, gender macro-F1 improved from 0.7223 to 0.8164 and direct age fine-tuning improved macro-F1 from 0.4410 to 0.5310. The final externally balanced age model achieved 0.5348 macro-F1 on the original target-domain test set and 0.7587 on a separate class-balanced, mixed-source test set. These results show that evaluation source and class distribution materially affect age classification performance.
- 18 September 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 153 (1) , 9
Open AccessImproving Pedestrian Age and Gender Classification in a DeepStream-Based Video Analytics Pipeline
Real-time pedestrian analytics systems often combine detection, tracking, and demographic classification. However, generic demographic classifiers may generalize poorly to deployment-specific camera conditions. This paper presents a DeepStream-based pipeline using PeopleNet, object tracking, and independent ResNet18 classifiers for gender and age estimation. A target-domain dataset of 3508 labeled pedestrian crops was used for fine-tuning. The gender model used validation-based threshold calibration, while age optimization included imbalance-aware training and external data balancing. On the target-domain test set, gender macro-F1 improved from 0.7223 to 0.8164 and direct age fine-tuning improved macro-F1 from 0.4410 to 0.5310. The final externally balanced age model achieved 0.5348 macro-F1 on the original target-domain test set and 0.7587 on a separate class-balanced, mixed-source test set. These results show that evaluation source and class distribution materially affect age classification performance.
- Open Access
- 18 September 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 153 (1) , 36
A Computer Vision Framework for Galápagos Bird Species Identification: Dataset, CNN Benchmarking, and Web/Mobile Deployment
We present the first curated dataset and deployable web/mobile prototype dedicated to the automatic identification of bird species recorded for the Galápagos Islands, a UNESCO World Heritage Site of exceptional ecological importance. The dataset was constructed through a multi-stage pipeline—automated retrieval, filtering, and checklist-guided multi-reviewer validation—using the Charles Darwin Foundation catalogue as a taxonomic reference. It comprises 6492 high-quality images covering 58 species, and represents the first publicly available resource specifically tailored to Galápagos checklist bird species. Leveraging this dataset, we benchmarked several state-of-the-art convolutional neural networks, including VGG16, ResNet50, EfficientNetB0, InceptionV3, MobileNetV2, and ConvNeXt, thereby establishing a reference baseline for future research. ConvNeXt obtained the best performance, exceeding 95% aggregate top-1 accuracy in the class-stratified image-level evaluation. Based on this model, we developed a web and mobile application that supports interactive species identification directly through the device’s camera. The system adapts content to three user profiles—children, tourists, and scientists—ensuring both accessibility and scientific rigor. This contribution combines the development of a dataset, benchmark, and deployable tool, aiming to foster biodiversity awareness, promote citizen science, and support conservation and educational initiatives in the Galápagos Islands.
- 18 September 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 153 (1) , 36
Open AccessA Computer Vision Framework for Galápagos Bird Species Identification: Dataset, CNN Benchmarking, and Web/Mobile Deployment
We present the first curated dataset and deployable web/mobile prototype dedicated to the automatic identification of bird species recorded for the Galápagos Islands, a UNESCO World Heritage Site of exceptional ecological importance. The dataset was constructed through a multi-stage pipeline—automated retrieval, filtering, and checklist-guided multi-reviewer validation—using the Charles Darwin Foundation catalogue as a taxonomic reference. It comprises 6492 high-quality images covering 58 species, and represents the first publicly available resource specifically tailored to Galápagos checklist bird species. Leveraging this dataset, we benchmarked several state-of-the-art convolutional neural networks, including VGG16, ResNet50, EfficientNetB0, InceptionV3, MobileNetV2, and ConvNeXt, thereby establishing a reference baseline for future research. ConvNeXt obtained the best performance, exceeding 95% aggregate top-1 accuracy in the class-stratified image-level evaluation. Based on this model, we developed a web and mobile application that supports interactive species identification directly through the device’s camera. The system adapts content to three user profiles—children, tourists, and scientists—ensuring both accessibility and scientific rigor. This contribution combines the development of a dataset, benchmark, and deployable tool, aiming to foster biodiversity awareness, promote citizen science, and support conservation and educational initiatives in the Galápagos Islands.
- 2 September 2026
- Published by Springer Nature
- p. 344-353
Automated Diagnosis of Diabetic Retinopathy and Age-Related Macular Degeneration from Fundus Images
- 2 September 2026
- Published by Springer Nature
- p. 344-353
Automated Diagnosis of Diabetic Retinopathy and Age-Related Macular Degeneration from Fundus Images
- Open Access
Lightweight Machine Learning-Based QoS Optimization for Multi-UAV Emergency Communications in FANETs
Flying Ad Hoc Networks (FANETs) composed of multiple unmanned aerial vehicles (UAVs) are a promising solution for emergency wireless communications when terrestrial infrastructure is unavailable. However, ensuring reliable Quality of Service (QoS) in these highly dynamic networks remains challenging due to topology changes, varying propagation conditions, and congestion. This work proposes a lightweight machine learning-based QoS optimization framework for multi-UAV emergency communications that combines realistic mobility modeling, empirical channel measurements, and adaptive traffic prioritization. UAV mobility patterns are generated with ArduSim, while LoS/NLoS propagation models are derived from real UAV flight experiments and integrated into ns-3. Multiple supervised machine learning algorithms—including Decision Trees, Random Forest, Support Vector Machines, k-NN, Gradient Boosting, and CatBoost—are trained using four input features derived from the network state: , , , and . Simulation results show that the proposed AI SMOTE EMERGENCY scheme, based on CatBoost, improves the Packet Delivery Ratio (PDR) by approximately 43% over the No-QoS baseline, achieving 89–93% delivery across all four application ports. Compared with EDCA, the proposed scheme maintains reliable delivery for all services, increases emergency throughput by 34–36%, and reduces end-to-end delay by about 70%. In addition, the higher delivery reliability translates into clear communication energy benefits, reducing energy waste across all evaluated topologies when compared with the No-QoS baseline. The inference time remains below 0.002 s, supporting real-time QoS adaptation in resource-constrained UAV networks.Open AccessLightweight Machine Learning-Based QoS Optimization for Multi-UAV Emergency Communications in FANETs
Flying Ad Hoc Networks (FANETs) composed of multiple unmanned aerial vehicles (UAVs) are a promising solution for emergency wireless communications when terrestrial infrastructure is unavailable. However, ensuring reliable Quality of Service (QoS) in these highly dynamic networks remains challenging due to topology changes, varying propagation conditions, and congestion. This work proposes a lightweight machine learning-based QoS optimization framework for multi-UAV emergency communications that combines realistic mobility modeling, empirical channel measurements, and adaptive traffic prioritization. UAV mobility patterns are generated with ArduSim, while LoS/NLoS propagation models are derived from real UAV flight experiments and integrated into ns-3. Multiple supervised machine learning algorithms—including Decision Trees, Random Forest, Support Vector Machines, k-NN, Gradient Boosting, and CatBoost—are trained using four input features derived from the network state: , , , and . Simulation results show that the proposed AI SMOTE EMERGENCY scheme, based on CatBoost, improves the Packet Delivery Ratio (PDR) by approximately 43% over the No-QoS baseline, achieving 89–93% delivery across all four application ports. Compared with EDCA, the proposed scheme maintains reliable delivery for all services, increases emergency throughput by 34–36%, and reduces end-to-end delay by about 70%. In addition, the higher delivery reliability translates into clear communication energy benefits, reducing energy waste across all evaluated topologies when compared with the No-QoS baseline. The inference time remains below 0.002 s, supporting real-time QoS adaptation in resource-constrained UAV networks. - Open Access
- 16 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 179759-179775
Legal AI for All: Reducing Perplexity and Boosting Accuracy in Normative Texts With Fine-Tuned LLMs and RAG
This study investigates the potential of open-source Large Language Models (LLMs) to enhance the accessibility and understanding of complex legal information. Legal texts often present linguistic and structural barriers for non-experts, highlighting the need for tools that can support natural language interaction with normative content. We propose a framework that integrates Fine-Tuning and Retrieval-Augmented Generation (RAG) techniques to adapt LLMs for domain-specific legal applications. Using synthetic, validated question-answer datasets, we train and evaluate multiple open-source models and introduce novel benchmarks to assess performance across different difficulty levels. Results show that the fine-tuned models significantly reduce perplexity (up to 10×) and, when combined with RAG, achieve accuracy levels exceeding 90%. To demonstrate adaptability, an additional use case involving travel-related legal norms in an environmentally protected insular region is included. This confirms the scalability of the approach to diverse legal scenarios, supporting its potential for broader adoption in legal information systems.
- 16 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 179759-179775
Open AccessLegal AI for All: Reducing Perplexity and Boosting Accuracy in Normative Texts With Fine-Tuned LLMs and RAG
This study investigates the potential of open-source Large Language Models (LLMs) to enhance the accessibility and understanding of complex legal information. Legal texts often present linguistic and structural barriers for non-experts, highlighting the need for tools that can support natural language interaction with normative content. We propose a framework that integrates Fine-Tuning and Retrieval-Augmented Generation (RAG) techniques to adapt LLMs for domain-specific legal applications. Using synthetic, validated question-answer datasets, we train and evaluate multiple open-source models and introduce novel benchmarks to assess performance across different difficulty levels. Results show that the fine-tuned models significantly reduce perplexity (up to 10×) and, when combined with RAG, achieve accuracy levels exceeding 90%. To demonstrate adaptability, an additional use case involving travel-related legal norms in an environmentally protected insular region is included. This confirms the scalability of the approach to diverse legal scenarios, supporting its potential for broader adoption in legal information systems.
- Open Access
- 1 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 60930-60943
Correlation Between Depression-Associated Genes and Cancer Types: Predicting Cancer Based on Mutation Frequencies
This study examines the potential interconnection between genetic mutations frequently observed in various types of cancer and depression, aiming to elucidate shared genetic factors that may influence the etiology and treatment of both conditions. Genomic profiles from cancer patients were extracted using cBioPortal to perform comparisons with a subset of individuals carrying genetic variants linked to depression. The analysis employed advanced methodologies, including HJ biplot K-means and DBSCAN clustering algorithms for pattern grouping in 2D. This process generated a dataset, enabling the training and testing of machine learning and deep learning classification models. Clustering results highlighted genes associated with depression and mutation patterns across different cancer types, such as skin, uterine, cervical, stomach, and prostate cancer. Furthermore, in classifying cancers with similar characteristics potentially related to depression, an F1 score of 0.95 or higher was achieved using random forest, KNN, and neural network models. The implementation details and source code for this research are available in https://github.com/Aptroide/Genes_Protocol, providing transparency and facilitating further studies.
- 1 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 60930-60943
Open AccessCorrelation Between Depression-Associated Genes and Cancer Types: Predicting Cancer Based on Mutation Frequencies
This study examines the potential interconnection between genetic mutations frequently observed in various types of cancer and depression, aiming to elucidate shared genetic factors that may influence the etiology and treatment of both conditions. Genomic profiles from cancer patients were extracted using cBioPortal to perform comparisons with a subset of individuals carrying genetic variants linked to depression. The analysis employed advanced methodologies, including HJ biplot K-means and DBSCAN clustering algorithms for pattern grouping in 2D. This process generated a dataset, enabling the training and testing of machine learning and deep learning classification models. Clustering results highlighted genes associated with depression and mutation patterns across different cancer types, such as skin, uterine, cervical, stomach, and prostate cancer. Furthermore, in classifying cancers with similar characteristics potentially related to depression, an F1 score of 0.95 or higher was achieved using random forest, KNN, and neural network models. The implementation details and source code for this research are available in https://github.com/Aptroide/Genes_Protocol, providing transparency and facilitating further studies.
- Open Access
- 20 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 59687-59703
Predictive Modeling for Fetal Health: A Comparative Study of PCA, LDA, and KPCA for Dimensionality Reduction
Pregnancy complications significantly impact maternal and fetal health, requiring accurate and timely diagnostic methods for life-saving interventions. Traditional manual analysis of cardiotocography (CTG) tests, a commonly used technique for fetal health monitoring, is labor-intensive and prone to variability. This study addresses this issue by applying dimensionality reduction techniques—Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Kernel Principal Component Analysis (KPCA)—to improve the performance of machine learning (ML) models in predicting fetal health using CTG data. Using a dataset of 2,126 records, our goal was to reduce the feature space while preserving essential information, resulting in improved classification accuracy. The results show that PCA with XGBoost achieved 98% accuracy, LDA with K-nearest Neighbors achieved 91%, and KPCA with XGBoost achieved 97%. These findings highlight the importance of dimensionality reduction and feature selection in developing robust ML models for fetal health assessment, emphasizing their potential impact on improving clinical diagnostics and medical decision-making.
- 20 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 59687-59703
Open AccessPredictive Modeling for Fetal Health: A Comparative Study of PCA, LDA, and KPCA for Dimensionality Reduction
Pregnancy complications significantly impact maternal and fetal health, requiring accurate and timely diagnostic methods for life-saving interventions. Traditional manual analysis of cardiotocography (CTG) tests, a commonly used technique for fetal health monitoring, is labor-intensive and prone to variability. This study addresses this issue by applying dimensionality reduction techniques—Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Kernel Principal Component Analysis (KPCA)—to improve the performance of machine learning (ML) models in predicting fetal health using CTG data. Using a dataset of 2,126 records, our goal was to reduce the feature space while preserving essential information, resulting in improved classification accuracy. The results show that PCA with XGBoost achieved 98% accuracy, LDA with K-nearest Neighbors achieved 91%, and KPCA with XGBoost achieved 97%. These findings highlight the importance of dimensionality reduction and feature selection in developing robust ML models for fetal health assessment, emphasizing their potential impact on improving clinical diagnostics and medical decision-making.