eprintid: 17573 rev_number: 8 eprint_status: archive userid: 2 dir: disk0/00/01/75/73 datestamp: 2025-04-09 23:30:12 lastmod: 2025-04-09 23:30:13 status_changed: 2025-04-09 23:30:12 type: article metadata_visibility: show creators_name: Naseer, Aisha creators_name: Amjad, Madiha creators_name: Raza, Ali creators_name: Munir, Kashif creators_name: Smerat, Aseel creators_name: Fabian Gongora, Henry creators_name: Uc Ríos, Carlos Eduardo creators_name: Ashraf, Imran creators_id: creators_id: creators_id: creators_id: creators_id: creators_id: henry.gongora@uneatlantico.es creators_id: carlos.uc@unini.edu.mx creators_id: title: Novel hybrid transfer neural network for wheat crop growth stages recognition using field images ispublished: pub subjects: uneat_eng divisions: uneatlantico_produccion_cientifica divisions: uninimx_produccion_cientifica divisions: uninipr_produccion_cientifica divisions: uniromana_produccion_cientifica full_text_status: public keywords: Precision agriculture, Agricultural system, Wheat growth prediction, Hybrid neural network, Image processing, Deep learning abstract: Wheat is one of the world’s most widely cultivated cereal crops and is a primary food source for a significant portion of the population. Wheat goes through several distinct developmental phases, and accurately identifying these stages is essential for precision farming. Determining wheat growth stages accurately is crucial for increasing the efficiency of agricultural yield in wheat farming. Preliminary research identified obstacles in distinguishing between these stages, negatively impacting crop yields. To address this, this study introduces an innovative approach, MobDenNet, based on data collection and real-time wheat crop stage recognition. The data collection utilized a diverse image dataset covering seven growth phases ‘Crown Root’, ‘Tillering’, ‘Mid Vegetative’, ‘Booting’, ‘Heading’, ‘Anthesis’, and ‘Milking’, comprising 4496 images. The collected image dataset underwent rigorous preprocessing and advanced data augmentation to refine and minimize biases. This study employed deep and transfer learning models, including MobileNetV2, DenseNet-121, NASNet-Large, InceptionV3, and a convolutional neural network (CNN) for performance comparison. Experimental evaluations demonstrated that the transfer model MobileNetV2 achieved 95% accuracy, DenseNet-121 achieved 94% accuracy, NASNet-Large achieved 76% accuracy, InceptionV3 achieved 74% accuracy, and the CNN achieved 68% accuracy. The proposed novel hybrid approach, MobDenNet, that synergistically merges the architectures of MobileNetV2 and DenseNet-121 neural networks, yields highly accurate results with precision, recall, and an F1 score of 99%. We validated the robustness of the proposed approach using the k-fold cross-validation. The proposed research ensures the detection of growth stages with great promise for boosting agricultural productivity and management practices, empowering farmers to optimize resource distribution and make informed decisions. date: 2025-04 publication: Scientific Reports volume: 15 number: 1 id_number: doi:10.1038/s41598-025-96332-9 refereed: TRUE issn: 2045-2322 official_url: http://doi.org/10.1038/s41598-025-96332-9 access: open language: en citation: Artículo Materias > Ingeniería Universidad Europea del Atlántico > Investigación > Producción Científica Universidad Internacional Iberoamericana México > Investigación > Producción Científica Universidad Internacional Iberoamericana Puerto Rico > Investigación > Producción Científica Universidad de La Romana > Investigación > Producción Científica Abierto Inglés Wheat is one of the world’s most widely cultivated cereal crops and is a primary food source for a significant portion of the population. Wheat goes through several distinct developmental phases, and accurately identifying these stages is essential for precision farming. Determining wheat growth stages accurately is crucial for increasing the efficiency of agricultural yield in wheat farming. Preliminary research identified obstacles in distinguishing between these stages, negatively impacting crop yields. To address this, this study introduces an innovative approach, MobDenNet, based on data collection and real-time wheat crop stage recognition. The data collection utilized a diverse image dataset covering seven growth phases ‘Crown Root’, ‘Tillering’, ‘Mid Vegetative’, ‘Booting’, ‘Heading’, ‘Anthesis’, and ‘Milking’, comprising 4496 images. The collected image dataset underwent rigorous preprocessing and advanced data augmentation to refine and minimize biases. This study employed deep and transfer learning models, including MobileNetV2, DenseNet-121, NASNet-Large, InceptionV3, and a convolutional neural network (CNN) for performance comparison. Experimental evaluations demonstrated that the transfer model MobileNetV2 achieved 95% accuracy, DenseNet-121 achieved 94% accuracy, NASNet-Large achieved 76% accuracy, InceptionV3 achieved 74% accuracy, and the CNN achieved 68% accuracy. The proposed novel hybrid approach, MobDenNet, that synergistically merges the architectures of MobileNetV2 and DenseNet-121 neural networks, yields highly accurate results with precision, recall, and an F1 score of 99%. We validated the robustness of the proposed approach using the k-fold cross-validation. The proposed research ensures the detection of growth stages with great promise for boosting agricultural productivity and management practices, empowering farmers to optimize resource distribution and make informed decisions. metadata Naseer, Aisha; Amjad, Madiha; Raza, Ali; Munir, Kashif; Smerat, Aseel; Fabian Gongora, Henry; Uc Ríos, Carlos Eduardo y Ashraf, Imran mail SIN ESPECIFICAR, SIN ESPECIFICAR, SIN ESPECIFICAR, SIN ESPECIFICAR, SIN ESPECIFICAR, henry.gongora@uneatlantico.es, carlos.uc@unini.edu.mx, SIN ESPECIFICAR (2025) Novel hybrid transfer neural network for wheat crop growth stages recognition using field images. Scientific Reports, 15 (1). ISSN 2045-2322 document_url: http://repositorio.uniromana.edu.do/id/eprint/17573/1/s41598-025-96332-9.pdf