U2ESPNet—A lightweight and high-accuracy convolutional neural network for real-time semantic segmentation of visible branches

Wan, Hao; Zeng, Xilei; Zhao, Fan; Zhang, Shanshan; Kang, Meilin

doi:10.1016/j.compag.2022.107542

Cited by 10 publications

(3 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The F1 score of model IV reaches 0.812, which is higher than those of the other three models. Although the precision of model IV is not the highest, the recall and IoU of model IV are significantly higher, which can reflect the extent of capture of the target pixel [88]. The performance of the U-Net model is better than that of the SegNet model, regardless of the two different deformation indexes, which means that the U-Net model is more suitable for obtaining abstract information.…”

Section: Experimental Results 441 Comparison Of Different Deformation...mentioning

confidence: 95%

Automatic Identification for the Boundaries of InSAR Anomalous Deformation Areas Based on Semantic Segmentation Model

Liang,

Zhang,

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Interferometric synthetic aperture radar (InSAR) technology has become one of the mainstream techniques for active landslide identification over a large area. However, the method for interpreting anomalous deformation areas derived from InSAR data is still mainly manual delineation through human–computer interaction. This study focuses on using a deep learning semantic segmentation model to identify the boundaries of anomalous deformation areas automatically. We experimented with the delineation results based on an InSAR deformation map, hot spot map, and different combinations of topographic datasets to build the optimal model. The result indicates that the hot spot map, aspect, and Google Earth image as input features based on the U-Net model can achieve the best performance, with the precision, recall, F1 score, and intersection over union (IoU) being 0.822, 0.835, 0.823, and 0.705, respectively. Our method promotes the development of identifying active landslides using InSAR technology automatically and rapidly at a regional scale. Moreover, applying a new method for automatically and rapidly identifying potential landslides in susceptible areas is necessary for landslide hazard mitigation and risk management.

show abstract

Section: Experimental Results 441 Comparison Of Different Deformation...mentioning

confidence: 95%

Automatic Identification for the Boundaries of InSAR Anomalous Deformation Areas Based on Semantic Segmentation Model

Liang,

Zhang,

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…This notable achievement serves to underscore the model's efficacy in accurately recognizing maize growth stages. It could be observed that the segmentation accuracy of the original U-net model showed significant improvement due to the stable and efficient feature extraction ability of the cascaded convolutional structure [74]; the upgraded decoder with embedded CA module that focuses on connected regions while suppressing background feature expression [75]; and the dilation path that effectively preserves spatial information [76]. This finding indicated that the improved U-net model achieved optimal ac-curacy in recognition of the maize growth stage and obtained a satisfactory result, successfully recognizing the maize growth stage in most cases.…”

Section: Comparison With Other Semantic Segmentation Modelsmentioning

confidence: 99%

Improved U-Net for Growth Stage Recognition of In-Field Maize

et al. 2023

View full text Add to dashboard Cite

Precise recognition of maize growth stages in the field is one of the critical steps in conducting precision irrigation and crop growth evaluation. However, due to the ever-changing environmental factors and maize growth characteristics, traditional recognition methods usually suffer from limitations in recognizing different growth stages. For the purpose of tackling these issues, this study proposed an improved U-net by first using a cascade convolution-based network as the encoder with a strategy for backbone network replacement to optimize feature extraction and reuse. Secondly, three attention mechanism modules have been introduced to upgrade the decoder part of the original U-net, which highlighted critical regions and extracted more discriminative features of maize. Subsequently, a dilation path of the improved U-net was constructed by integrating dilated convolution layers using a multi-scale feature fusion approach to preserve the detailed spatial information of in-field maize. Finally, the improved U-net has been applied to recognize different growth stages of maize in the field. The results clearly demonstrated the superior ability of the improved U-net to precisely segment and recognize maize growth stage from in-field images. Specifically, the semantic segmentation network achieved a mean intersection over union (mIoU) of 94.51% and a mean pixel accuracy (mPA) of 96.93% in recognizing the maize growth stage with only 39.08 MB of parameters. In conclusion, the good trade-offs made in terms of accuracy and parameter number demonstrated that this study could lay a good foundation for implementing accurate maize growth stage recognition and long-term automatic growth monitoring.

show abstract

“…Em Hao et al [24] os autores propõem um modelo de CNN leve chamado U2ESPNet aplicado à segmentação de ramos de macieiras em pomares. Esse modelo visa melhorar a inteligência dos robôs de colheita de maçãs e possibilitar o reconhecimento de ramos com alta precisão e em tempo real, mesmo em dispositivos com recursos limitados.…”

Section: ____________________________________________________________...unclassified

Potencializando a classificação de imagens com eficiência e robustez através da FiberNet.

Rafael Ferreira,

Magaly de Paula Canuto

2023

ReCiC

View full text Add to dashboard Cite

Um modelo baseado em CNN que utiliza umpequeno número de parâmetros treináveis pode trazer váriosbenefícios importantes para diferentes domínios deaplicação. Entre eles, podemos mencionar o processo detreinamento mais rápido, o que significa que suaimplementação pode ser feita de forma mais eficiente erápida. Além deste, nós podemos citar também a reduçãosignificativa na probabilidade de overfitting e a redução noconsumo dos recursos computacionais necessários para aexecução, o que pode ser muito útil em sistemas embarcadose outros dispositivos com hardware limitado. Portanto, nesteartigo, nós propomos o modelo FiberNet, que é uma redeneural convolucional (CNN) simples e eficiente que produzum pequeno número de parâmetros treináveis e altavelocidade de inferência. O objetivo principal é fornecer umaalternativa eficiente para aplicações móveis, bem comoaplicações que requerem modelos compactos. Para avaliar aeficácia da FiberNet, nós realizamos um estudo empíricocomparativo entre os resultados do nosso modelo e oresultado de outros modelos que são considerados como oestado da arte. Essa comparação foi realizada em doisconjuntos de dados de imagens diferentes, um dos quais é oconhecido conjunto de dados CIFAR10. Como resultado, omodelo FiberNet obteve uma precisão de 96,25% noconjunto de dados Sisal, 74,90% no conjunto de dadosCIFAR10 e foi o primeiro em total de parâmetros treináveiscom 754.345 (39,57% menor em relação ao segundocolocado)

show abstract

U2ESPNet—A lightweight and high-accuracy convolutional neural network for real-time semantic segmentation of visible branches

Cited by 10 publications

References 23 publications

Automatic Identification for the Boundaries of InSAR Anomalous Deformation Areas Based on Semantic Segmentation Model

Automatic Identification for the Boundaries of InSAR Anomalous Deformation Areas Based on Semantic Segmentation Model

Improved U-Net for Growth Stage Recognition of In-Field Maize

Potencializando a classificação de imagens com eficiência e robustez através da FiberNet.

Contact Info

Product

Resources

About