Weakly supervised attention-based models using activation maps for citrus mite and insect pest classification

Bollis, Edson; Maia, Helena de Almeida; Pedrini, Hélio; Avila, Sandra

doi:10.1016/j.compag.2022.106839

Cited by 16 publications

(9 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Therefore, computing the most important pixels in an input image is one way of determining the exact annotation. This strategy has been employed for object detection of crop pests ( Bollis et al., 2022 ) and segmentation of foliar diseases ( Yi et al., 2021 ). However, this challenge currently receives less attention than incomplete and inaccurate annotations.…”

Section: Imperfect Datasetmentioning

confidence: 99%

Embracing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning

Xu,

Kim,

Yang

et al. 2023

Front. Plant Sci.

View full text Add to dashboard Cite

Recent advancements in deep learning have brought significant improvements to plant disease recognition. However, achieving satisfactory performance often requires high-quality training datasets, which are challenging and expensive to collect. Consequently, the practical application of current deep learning–based methods in real-world scenarios is hindered by the scarcity of high-quality datasets. In this paper, we argue that embracing poor datasets is viable and aims to explicitly define the challenges associated with using these datasets. To delve into this topic, we analyze the characteristics of high-quality datasets, namely, large-scale images and desired annotation, and contrast them with the limited and imperfect nature of poor datasets. Challenges arise when the training datasets deviate from these characteristics. To provide a comprehensive understanding, we propose a novel and informative taxonomy that categorizes these challenges. Furthermore, we offer a brief overview of existing studies and approaches that address these challenges. We point out that our paper sheds light on the importance of embracing poor datasets, enhances the understanding of the associated challenges, and contributes to the ambitious objective of deploying deep learning in real-world applications. To facilitate the progress, we finally describe several outstanding questions and point out potential future directions. Although our primary focus is on plant disease recognition, we emphasize that the principles of embracing and analyzing poor datasets are applicable to a wider range of domains, including agriculture. Our project is public available at https://github.com/xml94/EmbracingLimitedImperfectTrainingDatasets.

show abstract

Section: Imperfect Datasetmentioning

confidence: 99%

Embracing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning

Xu,

Kim,

Yang

et al. 2023

Front. Plant Sci.

View full text Add to dashboard Cite

show abstract

“…Some mature CNN architectures have integrated attention modules, such as the SE module added to MobileNet V3, making it perform better than MobileNet V1 and MobileNet V2 [33]. In the field of agrofood, more and more researchers add attention modules to their self-designed network architectures to deal with challenges like complex backgrounds and inter-class similarity, such as leaf disease detection [51][52][53][54] and pest detection [55]. However, the attention modules are usually built by a series of complex factors, e.g., the choice for pooling.…”

Section: Attention Mechanismmentioning

confidence: 99%

A cross-domain fruit classification method based on lightweight attention networks and unsupervised domain adaptation

Wang

Zhang

Yan

et al. 2022

Complex Intell. Syst.

View full text Add to dashboard Cite

Image-based fruit classification offers many useful applications in industrial production and daily life, such as self-checkout in the supermarket, automatic fruit sorting and dietary guidance. However, fruit classification task will have different data distributions due to different application scenarios. One feasible solution to solve this problem is to use domain adaptation that adapts knowledge from the original training data (source domain) to the new testing data (target domain). In this paper, we propose a novel deep learning-based unsupervised domain adaptation method for cross-domain fruit classification. A hybrid attention module is proposed and added to MobileNet V3 to construct the HAM-MobileNet that can suppress the impact of complex backgrounds and extract more discriminative features. A hybrid loss function combining subdomain alignment and implicit distribution metrics is used to reduce domain discrepancy during model training and improve model classification performance. Two fruit classification datasets covering several domains are established to simulate common industrial and daily life application scenarios. We validate the proposed method on our constructed grape classification dataset and general fruit classification dataset. The experimental results show that the proposed method achieves an average accuracy of 95.0% and 93.2% on the two datasets, respectively. The classification model after domain adaptation can well overcome the domain discrepancy brought by different fruit classification scenarios. Meanwhile, the proposed datasets and method can serve as a benchmark for future cross-domain fruit classification research.

show abstract

“…Secondly, weakly supervised approaches may incur additional computational costs due to the use of complex architectures for learning the relationship between features and locations 32 or the generation of multiple heatmaps at different scales for accurate object localization 33 . Lastly, some weakly supervised approaches may involve a two-stage process that relies on pre-trained models for initial localization, potentially limiting the model's ability to directly learn from the heatmaps used for refinement.…”

Section: Introductionmentioning

confidence: 99%

HAFREE: A heatmap-based anchor-free detector for apple defects

Kim,

Nguyen,

Lee

2024

Preprint

View full text Add to dashboard Cite

Precise inspection of subtle defects on apple surfaces is crucial in real-world agricultural scenarios. However, existing methods often rely on expensive equipment or struggle to detect visually ambiguous defects. To address this challenge, we introduce the Surface Subtle Defects Apple (SSDA) dataset, a custom dataset collected and annotated specifically for this study. Using the SSDA dataset, we propose HAFREE (Heatmap-based Anchor-Free detector), a novel end-to-end object detection architecture designed for localizing subtle defects on apple surfaces. Unlike traditional methods reliant on anchor boxes, our model leverages a heatmap-based detector that represents defects as keypoints using 2D Gaussian kernels. To address the scarcity of labeled data in real-world settings, we propose a patch training strategy that incorporates both full images and cropped patches during training, providing the network with both local and global contextual information. Furthermore, we introduce a Multi-scale Feature Fusion Block (MFFB) to enrich visual patterns and enhance overall performance. Our method achieves an mAP@.5 of 50.05% on the SSDA dataset, outperforming both two-stage and one-stage anchor-based detectors and state-of-the-art anchor-free approaches. HAFREE paves the way for a new paradigm in agricultural defect detection, offering an effective precise solution for real-world applications.

show abstract

Weakly supervised attention-based models using activation maps for citrus mite and insect pest classification

Cited by 16 publications

References 21 publications

Embracing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning

Embracing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning

A cross-domain fruit classification method based on lightweight attention networks and unsupervised domain adaptation

HAFREE: A heatmap-based anchor-free detector for apple defects

Contact Info

Product

Resources

About