Agri-PIDNet-S: Lightweight semantic segmentation for field-structure and obstacle perception in seedling-stage cotton fields
- 1College of Computer and Information Engineering, Xinjiang Agricultural University, Urumqi, China.
- 2Engineering Research Center of Intelligent Agriculture, Ministry of Education, Urumqi, China.
- 3Xinjiang Agricultural Informatization Engineering Technology Research Center, Urumqi, China.
Abstract
Reliable joint perception of crop rows, furrows, and obstacles is essential for autonomous inter-row operations in seedling-stage cotton fields. However, small and sparsely distributed plants, perspective-convergent rows, irregular furrow boundaries, and low obstacle pixel coverage jointly challenge fine-detail preservation, long-range structural modeling, and boundary discrimination under real-time constraints. This study formulated cotton-field perception as a four-class semantic segmentation task and constructed a dataset of 709 pixel-wise annotated images comprising background, crop rows, furrows, and obstacles. We developed Agri-PIDNet-S, a field-specific lightweight network based on PIDNet-S. A coordinate-aware multi-kernel inverted residual module (CA-MKIR) preserves shallow details and small-obstacle contours, a squeeze-and-excitation-enhanced deep-to-token module (SaE-D2T) models long-range row and furrow structures, and a boundary-guided window-attention refinement Bag (BWR-Bag) improves semantic–boundary fusion. The network was trained with a composite loss incorporating auxiliary boundary supervision. On the cotton-field test set, Agri-PIDNet-S achieved 92.37% mean intersection over union (mIoU) and 96.02% mean Dice coefficient (mDice) in the model-comparison experiment, while repeated training with three seeds yielded an mIoU of 92.35 ± 0.03%. Relative to PIDNet-S under the same comparison protocol, mIoU increased by 3.53 percentage points, while obstacle IoU improved from 79.60% to 88.67%. The model contained 8.048 million parameters, required 19.161 GFLOPs, and achieved 155.67 frames per second. Zero-shot external evaluation on tomato, watermelon, and maize fields yielded a structural mIoU of 86.29% for the background, crop-row, and furrow classes. Under the evaluated conditions, the task-oriented modifications achieved higher segmentation point estimates than PIDNet-S while maintaining computational efficiency; the external results provide preliminary evidence of cross-crop structural transfer.
Related
- Frequency-specific oscillations track parallel processing of linguistic structural and lexical-semantic information from perception to decision
- Lightweight Semantic EEG Foundation Model for Frozen Cross-Disorder Transfer
- Dynamic Semantic–Prosodic Integration in Emotional Speech Perception: From Channel Dominance to Precision-Weighted Predictive Coding
- ACMNet: An attention-convolution mixed neural network for semantic segmentation of high-resolution remote sensing images
- Lightweight visual detection approach for apple ripeness perception in intelligent orchard harvesting scenarios
BibTeX
@article{xin2026agripidnets,
title = {{Agri-PIDNet-S: Lightweight semantic segmentation for field-structure and obstacle perception in seedling-stage cotton fields}},
author = {Yanlin Xin and Yongke Li and Wuchen Wei and Xinyu Cao and Bowen Mao and Yanhong Chen},
journal = {PLOS One},
year = {2026},
month = nov,
doi = {10.1371/journal.pone.0361992},
url = {https://doi.org/10.1371/journal.pone.0361992},
note = {CC BY},
}
RIS
TY - JOUR
TI - Agri-PIDNet-S: Lightweight semantic segmentation for field-structure and obstacle perception in seedling-stage cotton fields
AU - Yanlin Xin
AU - Yongke Li
AU - Wuchen Wei
AU - Xinyu Cao
AU - Bowen Mao
AU - Yanhong Chen
PY - 2026
DA - 2026/11/02
JO - PLOS One
DO - 10.1371/journal.pone.0361992
UR - https://doi.org/10.1371/journal.pone.0361992
AB - Reliable joint perception of crop rows, furrows, and obstacles is essential for autonomous inter-row operations in seedling-stage cotton fields. However, small and sparsely distributed plants, perspective-convergent rows, irregular furrow boundaries, and low obstacle pixel coverage jointly challenge fine-detail preservation, long-range structural modeling, and boundary discrimination under real-time constraints. This study formulated cotton-field perception as a four-class semantic segmentation task and constructed a dataset of 709 pixel-wise annotated images comprising background, crop rows, furrows, and obstacles. We developed Agri-PIDNet-S, a field-specific lightweight network based on PIDNet-S. A coordinate-aware multi-kernel inverted residual module (CA-MKIR) preserves shallow details and small-obstacle contours, a squeeze-and-excitation-enhanced deep-to-token module (SaE-D2T) models long-range row and furrow structures, and a boundary-guided window-attention refinement Bag (BWR-Bag) improves semantic–boundary fusion. The network was trained with a composite loss incorporating auxiliary boundary supervision. On the cotton-field test set, Agri-PIDNet-S achieved 92.37% mean intersection over union (mIoU) and 96.02% mean Dice coefficient (mDice) in the model-comparison experiment, while repeated training with three seeds yielded an mIoU of 92.35 ± 0.03%. Relative to PIDNet-S under the same comparison protocol, mIoU increased by 3.53 percentage points, while obstacle IoU improved from 79.60% to 88.67%. The model contained 8.048 million parameters, required 19.161 GFLOPs, and achieved 155.67 frames per second. Zero-shot external evaluation on tomato, watermelon, and maize fields yielded a structural mIoU of 86.29% for the background, crop-row, and furrow classes. Under the evaluated conditions, the task-oriented modifications achieved higher segmentation point estimates than PIDNet-S while maintaining computational efficiency; the external results provide preliminary evidence of cross-crop structural transfer.
SN - 1932-6203
ER -