Performance Evaluation of Semi-supervised Learning Frameworks for Multi-Class Weed Detection

TL;DR

This work tackles the challenge of labeling-intensive multi-class weed detection by evaluating a semi-supervised, teacher–student framework applied to both anchor-based (Faster-RCNN) and anchor-free (FCOS) detectors. By leveraging unlabeled data through improved pseudo-labeling and an unsupervised regression loss, the approach demonstrates substantial performance gains over supervised baselines on two cotton weed datasets (CottonWeedDet3 and CottonWeedDet12), achieving near-fully-supervised accuracy with only a fraction of labeled data. Key contributions include cross-detector evaluation, a generalized pseudo-labeling strategy that works for both detector types, and a publicly released codebase to foster ongoing research in label-efficient weed detection. The findings suggest that semi-supervised learning can meaningfully reduce labeling labor while maintaining or surpassing performance, supporting practical deployment in precision agriculture.

Abstract

Effective weed control plays a crucial role in optimizing crop yield and enhancing agricultural product quality. However, the reliance on herbicide application not only poses a critical threat to the environment but also promotes the emergence of resistant weeds. Fortunately, recent advances in precision weed management enabled by ML and DL provide a sustainable alternative. Despite great progress, existing algorithms are mainly developed based on supervised learning approaches, which typically demand large-scale datasets with manual-labeled annotations, which is time-consuming and labor-intensive. As such, label-efficient learning methods, especially semi-supervised learning, have gained increased attention in the broader domain of computer vision and have demonstrated promising performance. These methods aim to utilize a small number of labeled data samples along with a great number of unlabeled samples to develop high-performing models comparable to the supervised learning counterpart trained on a large amount of labeled data samples. In this study, we assess the effectiveness of a semi-supervised learning framework for multi-class weed detection, employing two well-known object detection frameworks, namely FCOS and Faster-RCNN. Specifically, we evaluate a generalized student-teacher framework with an improved pseudo-label generation module to produce reliable pseudo-labels for the unlabeled data. To enhance generalization, an ensemble student network is employed to facilitate the training process. Experimental results show that the proposed approach is able to achieve approximately 76\% and 96\% detection accuracy as the supervised methods with only 10\% of labeled data in CottenWeedDet3 and CottonWeedDet12, respectively. We offer access to the source code, contributing a valuable resource for ongoing semi-supervised learning research in weed detection and beyond.

Figures (7)

• Figure 1: Weed samples in the CottonWeedDet3 dataset rahman2023performance. Each column represents the image samples for one weed class.
• Figure 2: Weed samples in the CottonWeedDet12 dataset dang2023yoloweeds.
• Figure 3: Pipeline of the proposed semi-supervised weed detection framework.
• Figure 4: Training curves for FCOS and Faster RCNN with different proportions of labeled samples for two cotton weed datasets: CottonWeedDet3 (left) and CottonWeedDet12 (right).
• Figure 5: Examples of images annotated with ground truth labels (a) and predicted labels (b) using semi-supervised FOCS for CottonWeedDet3.