YOLOF-Snake: An Efficient Segmentation Model for Green Object Fruit

Jia, Weikuan; Liu, Mengyuan; Luo, Rong; Wang, Chongjing; Pan, Ningning; Yang, Xiaoli; Ge, Xinting

doi:10.3389/fpls.2022.765523

Cited by 8 publications

(1 citation statement)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The YOLO family of target detection algorithms has been applied to many fruit studies, such as green apple [10], citrus [11], strawberry [12], tomato [13], and pear [14]. [15] used an improved YOLOv5 to detect tomato ripeness by replacing the backbone network with MobileNetV3 and conducting channel pruning to achieve model lightweight.…”

Section: Introductionmentioning

confidence: 99%

Young Fruits of "Okubo" Peach Recognition in Complex Environments Based on an Improved YOLOv8s

Jing,

Zhang,

Sun

et al. 2024

Preprint

View full text Add to dashboard Cite

To address the challenges of high similarity in height between young fruits and leaves, small size of fruits, dense distribution, and severe occlusions, this paper proposes a lightweight YOLOv8s-P detection model for the identification of young fruits of "Okubo" peaches in complex environments. Firstly, the lightweight C2f_Faster module is designed and replaces all the C2f modules in YOLOv8s to realize the model lightweight. Secondly, the Efficient Multi-Scale Attention Module(EMA) is added inside the C2f_Faster module of the lightweight model to enhance the network's ability to extract tiny features. Finally, the MPDIoU boundary loss function based on the minimum point is used to replace the original CIoU boundary loss function of YOLOv8s to improve the positioning accuracy of the model prediction box. The results demonstrate that the YOLOv8s-P model achieves an average precision (AP@0.5) of 90.86%, F1 score of 86.70%, while only occupying 75.23% of the size of YOLOv8s. Compared to other lightweight algorithms such as YOLOv3-tiny, YOLOv4-tiny, YOLOv5s, YOLOv6s, YOLOv7-tiny, and YOLOv8s the YOLOv8s-P model achieves higher AP@0.5 by 6.26%, 6.01%, 2.05%, 2.12%, 1.87%, and 1.85% respectively. Additionally, the F1 score is higher by 3.93%, 3.42%, 1.54%, 1.52%, 0.68%, and 0.85% respectively. In conclusion, the YOLOv8s-P model has higher detection accuracy, compressed model size, and reduced hardware equipment configuration requirements, which provides a reference for the subsequent deployment and application of the "Okubo" peach fruit thinning robot hardware equipment.

show abstract

Section: Introductionmentioning

confidence: 99%

Young Fruits of "Okubo" Peach Recognition in Complex Environments Based on an Improved YOLOv8s

Jing,

Zhang,

Sun

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

Green fruit detection methods: Innovative application of camouflage object detection and multilevel feature mining

Zhai,

Gao,

Zhou

et al. 2024

Computers and Electronics in Agriculture

View full text Add to dashboard Cite

Apple detection and instance segmentation in natural environments using an improved Mask Scoring R-CNN Model

Wang

2022

Front. Plant Sci.

View full text Add to dashboard Cite

The accurate detection and segmentation of apples during growth stage is essential for yield estimation, timely harvesting, and retrieving growth information. However, factors such as the uncertain illumination, overlaps and occlusions of apples, homochromatic background and the gradual change in the ground color of apples from green to red, bring great challenges to the detection and segmentation of apples. To solve these problems, this study proposed an improved Mask Scoring region-based convolutional neural network (Mask Scoring R-CNN), known as MS-ADS, for accurate apple detection and instance segmentation in a natural environment. First, the ResNeSt, a variant of ResNet, combined with a feature pyramid network was used as backbone network to improve the feature extraction ability. Second, high-level architectures including R-CNN head and mask head were modified to improve the utilization of high-level features. Convolutional layers were added to the original R-CNN head to improve the accuracy of bounding box detection (bbox_mAP), and the Dual Attention Network was added to the original mask head to improve the accuracy of instance segmentation (mask_mAP). The experimental results showed that the proposed MS-ADS model effectively detected and segmented apples under various conditions, such as apples occluded by branches, leaves and other apples, apples with different ground colors and shadows, and apples divided into parts by branches and petioles. The recall, precision, false detection rate, and F1 score were 97.4%, 96.5%, 3.5%, and 96.9%, respectively. A bbox_mAP and mask_mAP of 0.932 and 0.920, respectively, were achieved on the test set, and the average run-time was 0.27 s per image. The experimental results indicated that the MS-ADS method detected and segmented apples in the orchard robustly and accurately with real-time performance. This study lays a foundation for follow-up work, such as yield estimation, harvesting, and automatic and long-term acquisition of apple growth information.

show abstract

YOLOF-Snake: An Efficient Segmentation Model for Green Object Fruit

Cited by 8 publications

References 40 publications

Young Fruits of "Okubo" Peach Recognition in Complex Environments Based on an Improved YOLOv8s

Young Fruits of "Okubo" Peach Recognition in Complex Environments Based on an Improved YOLOv8s

Green fruit detection methods: Innovative application of camouflage object detection and multilevel feature mining

Apple detection and instance segmentation in natural environments using an improved Mask Scoring R-CNN Model

Contact Info

Product

Resources

About