Few-shot learning approach with multi-scale feature fusion and attention for plant disease recognition

Abstract

Image-based deep learning method for plant disease diagnosing is promising but relies on large-scale dataset. Currently, the shortage of data has become an obstacle to leverage deep learning methods. Few-shot learning can generalize to new categories with the supports of few samples, which is very helpful for those plant disease categories where only few samples are available. However, two challenging problems are existing in few-shot learning: (1) the feature extracted from few shots is very limited; (2) generalizing to new categories, especially to another domain is very tough. In response to the two issues, we propose a network based on the Meta-Baseline few-shot learning method, and combine cascaded multi-scale features and channel attention. The network takes advantage of multi-scale features to rich the feature representation, uses channel attention as a compensation module efficiently to learn more from the significant channels of the fused features. Meanwhile, we propose a group of training strategies from data configuration perspective to match various generalization requirements. Through extensive experiments, it is verified that the combination of multi-scale feature fusion and channel attention can alleviate the problem of limited features caused by few shots. To imitate different generalization scenarios, we set different data settings and suggest the optimal training strategies for intra-domain case and cross-domain case, respectively. The effects of important factors in few-shot learning paradigm are analyzed. With the optimal configuration, the accuracy of 1-shot task and 5-shot task achieve at 61.24% and 77.43% respectively in the task targeting to single-plant, and achieve at 82.52% and 92.83% in the task targeting to multi-plants. Our results outperform the existing related works. It demonstrates that the few-shot learning is a feasible potential solution for plant disease recognition in the future application.

Plant and Disease Recognition Based on PMF Pipeline Domain Adaptation Method: Using Bark Images as Meta-Dataset

“…Compared to previous studies [ 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 ], our experiments achieved similar or even better results in terms of (1) the accuracy of prediction, despite being trained from bark, our tests on some public datasets (such as Agricultural Disease, PlantVillage, and Flowers) yielded promising results, with an average 5-shot accuracy of about 93%; (2) the ability of domain adaptation; while other methods may rely on more specific or domain-dependent features, our method can adapt to different regions, environments, and seasons more effectively than other methods; (3) the amount of data required (e.g., BarkVN50 has only 4000 images), reduce the cost and time for data collection and annotation; and (4) the transfer capability, as shown in the t-SNE visualization, the performance of the model is more stable in the transfer between domains. It is important to note that, unlike previous studies on FSL in agriculture, our work focuses on CDFSL.…”

“…Most of the studies have been conducted by exploiting different feature extraction, data augmentation, metric learning, and self-supervised training strategies to improve the accuracy, robustness, and generalization of FSL models. These methods have been applied in both spatial and frequency domains, covering image classification and target detection tasks [ 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 ]. Chen et al used a meta-learning framework for the adaptive process of the FSL plant disease detection task.…”

Few-shot disease recognition algorithm based on supervised contrastive learning

“…(2021) presented a few-shot method for detecting plant diseases called LFM-CNAPS, and the result showed that the model could detect unseen plant diseases using only 25 annotated examples with the average accuracy of 93.9%. Lin et al. (2022a) proposed a network based on the meta-baseline few-shot learning method, and combined the cascaded multi-scale features with channel attention.…”

A few-shot learning method for tobacco abnormality identification

“…The images of the same species always share many common features, which makes the classification difficult. That is the reason that in (Lin et al, 2022a, Lin et al, 2022b, the identification of the 10 categories of tomatoes is the most difficult task. For cases like task 1, an independent and static embedding without contextualization is not enough.…”