Saleh, Alzayat, Olsen, Alex, Wood, Jake, Philippa, Bronson, and Rahimi Azghadi, Mostafa (2025) Semi-supervised weed detection for rapid deployment and enhanced efficiency. Computers and Electronics in Agriculture, 236. 110410.

Preview

PDF (Published Version) - Published Version Available under License Creative Commons Attribution. Download (5MB) | Preview

1

Abstract

Weeds present a significant challenge in agriculture, causing yield loss and requiring expensive control measures. Automatic weed detection using computer vision and deep learning offers a promising solution. However, conventional deep learning methods often require large amounts of labeled training data, which can be costly and time-consuming to acquire. This paper introduces a novel method for semi-supervised weed detection, Semi-Supervised Multi-Scale Detector (SSMD), comprising two main components. Firstly, a multi-scale feature representation technique is employed to capture distinctive weed features across different scales. Secondly, the study proposes an adaptive pseudo-label assignment strategy, leveraging a small set of labeled images during training. This strategy dynamically assigns confidence scores to pseudo-labels generated from unlabeled data. Additionally, the proposed approach integrates epoch-corresponding and mixed pseudo-labels to further enhance the learning process. Experimental results on the COCO dataset and five prominent weed datasets, CottonWeedDet12, CropAndWeed, Palmer amaranth, RadishWheat, and RoboWeedMap, illustrate that the proposed SSMD achieves state-of-the-art performance in weed detection, even with significantly less labeled data compared to existing techniques. This SSMD holds the potential to alleviate the labeling burden and enhance the feasibility and deployment speed of deep learning for weed detection in real-world agricultural scenarios. The contributions of the proposed SSMD include: (1) Reduced Labeling Burden: The proposed approach significantly reduces the need for large amounts of labeled data, making deep learning models more practical and cost-effective for real-world deployments. (2) Improved Weed Detection Performance: Experiments demonstrate that the proposed method achieves state-of-the-art performance in weed detection with limited labeled data. (3) Enhanced Efficiency for Weed Management: The proposed method offers improved efficiency and accuracy, leading to better resource management and reduced environmental impact in agricultural applications.

Actions (Repository Staff Only)

Item Control Page