.png){kind=icon}
.png){kind=icon}
Return
Pocket-YOLO: Lightweight YOLO Model with Parameter-Free Attention for Peanut Stomata Detection
.png){kind=icon}
PRE
AI
Z
A
C
C
L
DOI：10.1007/s11760-026-05278-6.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Stomata are essential structures in terrestrial plants, and they are crucial for regulating photosynthesis and transpiration. The precise detection of stomata is essential for investigating traits like stomatal density and morphology. The traditional stomatal trait analysis methods through manual observation have the limitations of time-consuming, labor-intensive, and inefficient. In this paper, we propose a lightweight model named Pocket-YOLO to automatically identify stomata in peanut leaves. Specifically, we propose a structural Cross Stage Partial Network-Partial Convolution (CSPPC) to implement a lightweight model based on the Partial Convolution. The module effectively reduces computational redundancy and realizes the lightweight detection model. Additionally, the SimAM attention is integrated into the model to reduce the time-consuming of training phase and improve the detection accuracy. The convolutional layer in the YOLO model is optimized due to our integration of the CSPPC module. The number of model parameters is greatly reduced, and the computational GFLOPs significantly decrease as well. The SimAM module is added after the convolutional layer of the model head, which effectively improves the precision. Experimental results show that the Pocket-YOLO achieves a recognition accuracy of 94.6% on the peanut stomata dataset, which is a 0.5% increase compared to the YOLOv8 model. The Prediction accuracy is also significantly improved compared to the State-of-the-art model of YOLOv11. The proposed Pocket-YOLO enables complete automation of stomatal identification, and it is conducive to promoting large-scale research in agricultural science and botany.
Keywords:
Deep Learning
Stomata Detection
Attention Model
Pocket-YOLO
CSPPC
Journal
IF:
2.1
Papers:
778
Citations:
4.6K
Organization
Cited Papers