pith. machine review for the scientific record. sign in

arxiv: 2605.01283 · v1 · submitted 2026-05-02 · 💻 cs.CV · cs.AI

Recognition: unknown

Developing a Strong Pre-Trained Base Model for Plant Leaf Disease Classification

David J. Richter
Authors on Pith no claims yet

Pith reviewed 2026-05-09 14:48 UTC · model grok-4.3

classification 💻 cs.CV cs.AI
keywords plant leaf disease classificationtransfer learningDenseNet201pre-trained base modeldataset constructionimage augmentationCNNcrop disease detection
0
0 comments X

The pith

A DenseNet201 model pre-trained on a new combined plant leaf dataset outperforms general models as a starting point for transfer learning in disease classification.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper builds a new dataset by merging public plant leaf images and applying selected augmentations to train a specialized DenseNet201 model. This base model is then tested in transfer learning setups on additional disease datasets, where it produces classifiers that train faster, remain more stable, and succeed with smaller data amounts than those started from standard pre-trained networks. A sympathetic reader would care because reliable early detection of crop diseases can limit yearly yield losses without requiring extensive manual field inspections or large new training sets. The work shows that domain-focused dataset construction can close gaps left by available public resources and yield a reusable starting model for the task.

Core claim

The authors benchmark existing open datasets, run an augmentation study, and assemble a new training collection that supports training a DenseNet201 architecture. This resulting base model surpasses the performance of a standard baseline both on the new dataset itself and when used for transfer learning on a separate plant leaf disease collection, delivering training runs that are faster, more robust, more stable, and effective with less data than those begun from general pre-trained models.

What carries the argument

The new DenseNet201-based base model trained on the authors' custom dataset, which supplies a domain-specific initialization that improves transfer learning outcomes for plant leaf disease tasks.

If this is right

  • Transfer learning models derived from the new base require fewer labeled images to reach target accuracy on plant disease tasks.
  • Training runs converge faster and with greater stability than those using generic pre-trained networks.
  • The base model supports consistent performance across multiple plant species and disease types in the reported benchmarks.
  • Computational resources needed to develop new classifiers for leaf disease detection are reduced.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The dataset curation approach could be repeated for other narrow visual domains where public data falls short of what full models need.
  • Mobile or edge deployments of disease detection might become practical sooner if this base model reduces the data and compute barrier for fine-tuning.
  • Combining the base with additional sensor inputs or temporal sequences of field images could extend its usefulness for ongoing crop monitoring.

Load-bearing premise

The gains arise because the new dataset and augmentation choices produce a generally superior base model rather than one that benefits from dataset-specific biases or unstated differences in the benchmark setups.

What would settle it

A transfer learning experiment on a new plant disease dataset where a model started from this base model shows no improvement in training speed, stability, or data efficiency compared with one started from a standard general pre-trained network would challenge the central claim.

Figures

Figures reproduced from arXiv: 2605.01283 by David J. Richter.

Figure 1.1
Figure 1.1. Figure 1.1: Publications per year for ”deep learning AND plant leaf disease” papers based on view at source ↗
Figure 2.1
Figure 2.1. Figure 2.1: The frequency of datasets used in reviewed papers. view at source ↗
Figure 2.2
Figure 2.2. Figure 2.2: Frequency of used dataset types in papers review for this work. view at source ↗
Figure 2.3
Figure 2.3. Figure 2.3: Availability of datasets newly introduced in review papers. view at source ↗
Figure 2.4
Figure 2.4. Figure 2.4: The most used model families that were used in the review papers. view at source ↗
Figure 2
Figure 2. Figure 2: shows that among publicly downloaded datasets a total of 15 were taken in lab view at source ↗
Figure 2.5
Figure 2.5. Figure 2.5: The most used models in review work. most researchers worked with the same option (all though it comes with drawbacks), in model selection there seems to be no consensus as to which model or even model family to gravitate towards. A large number of different models from different architecture families see frequent use (see view at source ↗
Figure 3.1
Figure 3.1. Figure 3.1: This figure explains how kernels parse over the image and extract feature maps from them view at source ↗
Figure 3.2
Figure 3.2. Figure 3.2: Kernel operation visualized. 23 view at source ↗
Figure 3.3
Figure 3.3. Figure 3.3: Max Pooling process explained and visualized. 2 max pooling operations are shown view at source ↗
Figure 3.4
Figure 3.4. Figure 3.4: Flatten and Pooling visually explained view at source ↗
Figure 3.5
Figure 3.5. Figure 3.5: A fully connected neural network classifier top. single value per channel (see view at source ↗
Figure 3.6
Figure 3.6. Figure 3.6: Transfer-Learning method visualized scratch. This process uses a pre-trained model, a model that has previously been trained on a different dataset, and then trains that model, with the old weights still intact, to now work on the new domain specific data. Pre-trained models are usually trained on the ImageNet [58] dataset, a huge scale dataset with over 1.2 million images belonging to 1,000 classes. Thr… view at source ↗
Figure 3.7
Figure 3.7. Figure 3.7: Bacterial Diseases 3.2 Plant Pathology using Deep Learning These DL methods have been applied to many different fields in recent years and due to the ca￾pabilities of CNN to automatically learn features based on the labeled data, CNN are perfectly capable of being applied in many different fields successfully, granted the availability of suffi￾cient data. For the field of plant leaf disease classificatio… view at source ↗
Figure 3.8
Figure 3.8. Figure 3.8: Examples of Viral Diseases view at source ↗
Figure 3.9
Figure 3.9. Figure 3.9: Fungal Disease Examples 3.2.1.2 Viral Viral diseases can modify the DNA of plant host cells, which will make them produce new viruses, helping them reproduce. Viral diseases can be transmitted through water, but also via insects that move from one plant to the next [60]. Some viral diseases are mosaic, mop-top, etc. [60, 12] (see view at source ↗
Figure 4.1
Figure 4.1. Figure 4.1: Sample of Tomato Leaf images taken under Lab conditions from the Plant Village view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: Sample of Potato Leaf images taken taken in the view at source ↗
Figure 6.1
Figure 6.1. Figure 6.1: The workflow used during benchmarking 49 view at source ↗
Figure 7.1
Figure 7.1. Figure 7.1: The different color based data augmentation techniques used in this paper visual view at source ↗
Figure 7.2
Figure 7.2. Figure 7.2: The noise based data augmentation technique used in this paper visualized. Images view at source ↗
Figure 7.3
Figure 7.3. Figure 7.3: The different transformation based data augmentation techniques used in this paper view at source ↗
Figure 9.1
Figure 9.1. Figure 9.1: DenseNet201’s architecture before modification 68 view at source ↗
Figure 9.2
Figure 9.2. Figure 9.2: A visual comparison of the ReLU and the SiLU/Swish activation functions. view at source ↗
Figure 9.3
Figure 9.3. Figure 9.3: Flowchart of the Channel Attention Block. view at source ↗
Figure 9.4
Figure 9.4. Figure 9.4: The proposed PLDC-Net architecture based on the DenseNet201 baseline model view at source ↗
Figure 9.5
Figure 9.5. Figure 9.5: The result graph of the frozen TL results on the frozen 10-90 split. view at source ↗
Figure 9.6
Figure 9.6. Figure 9.6: The result graph of the frozen TL results on the frozen 30-70 split. view at source ↗
Figure 9.7
Figure 9.7. Figure 9.7: The result graph of the frozen TL results on the frozen 50-50 split. view at source ↗
Figure 9.8
Figure 9.8. Figure 9.8: The result graph of the frozen TL results on the frozen 70-30 split. view at source ↗
Figure 9.9
Figure 9.9. Figure 9.9: The result graph of the frozen TL results on the frozen 90-10 split. view at source ↗
Figure 9.10
Figure 9.10. Figure 9.10: The result graph of the unfrozen TL results on the unfrozen 10-90 split. view at source ↗
Figure 9.11
Figure 9.11. Figure 9.11: The zoomed in result graph of the unfrozen TL results on the unfrozen 10-90 split. view at source ↗
Figure 9.12
Figure 9.12. Figure 9.12: The result graph of the unfrozen TL results on the unfrozen 30-70 split. view at source ↗
Figure 9.13
Figure 9.13. Figure 9.13: The result graph of the unfrozen TL results on the unfrozen 50-50 split. view at source ↗
Figure 9.14
Figure 9.14. Figure 9.14: The result graph of the unfrozen TL results on the unfrozen 70-30 split. view at source ↗
Figure 9.15
Figure 9.15. Figure 9.15: The zoomed in result graph of the unfrozen TL results on the unfrozen 70-30 split. view at source ↗
Figure 9.16
Figure 9.16. Figure 9.16: The result graph of the unfrozen TL results on the unfrozen 90-10 split. view at source ↗
Figure 9.17
Figure 9.17. Figure 9.17: The zoomed in result graph of the unfrozen TL results on the unfrozen 90-10 split. view at source ↗
read the original abstract

Plants, crops and their yields are essential to our very existence, but diseases and pests cause large losses every year. As such it is vital to ensure that diseases can be spotted early and treated accordingly and stopping the spread while still possible. Manual and traditional methods require personal to walk through the field and check for symptoms 'by hand'. This is very laborious and very time consuming, so ML methods have been applied as a result and they have garnered promising results. CNN models are especially efficient as they can automatically extract features from images without any manual feature construction before then feeding the features to a classifier. Datasets are largely influential to the final performance of the model. Despite the importance that datasets pose to the field, there still seems to be somewhat of a discrepancy between what is publicly available for use and what would be required to sufficiently train fully capable models. To overcome these shortcomings, as part of this thesis open datasets for the field of plant leaf disease classification have been identified as well as models that can be trained on them and extensive benchmarks have been carried out to identify their suitability. Then a new dataset was constructed based on those findings as well as on the findings of a augmentation applicability study, which will be used to train a new Base Model based on the DenseNet201 architecture, which managed to outperform the baseline model on said new dataset as well as outperforming it on plant leaf disease classification domain specific Transfer-Learning experiments on another new dataset. This new model manages to train models through Transfer-Learning (TL) faster, more robust, more stable, and with less data than general model would, overcoming a large number of issues that the field still suffers from.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript identifies publicly available datasets for plant leaf disease classification, performs benchmarks to assess their suitability, constructs a new dataset informed by an augmentation study, trains a DenseNet201 base model on this dataset (claiming outperformance over baselines), and evaluates transfer learning on a second new dataset, asserting that the resulting model enables faster, more robust, stable, and data-efficient fine-tuning than ImageNet-pretrained general models.

Significance. If the empirical superiority claims hold under controlled conditions, the work could deliver a domain-adapted pre-trained model that mitigates data scarcity, training instability, and slow convergence issues common in CNN-based plant pathology tasks, providing a reusable base for downstream applications.

major comments (2)
  1. [Transfer learning experiments] Transfer learning experiments section: The claim that the new DenseNet201 base model trains faster, more robustly, more stably, and with less data than a general model requires explicit confirmation that fine-tuning protocols are identical between the new base model and the ImageNet baseline (same optimizer, learning-rate schedule, layer-freezing strategy, augmentation pipeline, batch size, and early-stopping rule). The abstract supplies no protocol details; any unstated differences would make the gains artifacts of setup rather than properties of the learned weights.
  2. [Results and evaluation] Results and evaluation sections: The abstract asserts outperformance on the new dataset and in TL experiments but supplies no concrete metrics, baselines, error bars, statistical tests, or validation protocols. This prevents independent verification of the central claims and undermines the data-efficiency and robustness assertions.
minor comments (2)
  1. [Abstract] Abstract: minor grammatical issues ('personal to walk through the field' should read 'personnel'; 'as part of this thesis' suggests the work may be thesis-derived and requires journal-style self-containment).
  2. [Dataset construction and methods] Notation and clarity: dataset names, augmentation choices, and exact baseline architectures are referenced but not tabulated or cross-referenced to specific sections, making reproducibility difficult.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We have addressed each major comment by clarifying the experimental setup and enhancing the reporting of results to improve verifiability. Revisions have been made to the relevant sections.

read point-by-point responses

  1. Referee: [Transfer learning experiments] Transfer learning experiments section: The claim that the new DenseNet201 base model trains faster, more robustly, more stably, and with less data than a general model requires explicit confirmation that fine-tuning protocols are identical between the new base model and the ImageNet baseline (same optimizer, learning-rate schedule, layer-freezing strategy, augmentation pipeline, batch size, and early-stopping rule). The abstract supplies no protocol details; any unstated differences would make the gains artifacts of setup rather than properties of the learned weights.

    Authors: We agree that identical protocols are essential for valid comparison. The fine-tuning protocols were the same for the domain-specific DenseNet201 and the ImageNet-pretrained baseline. We have revised the Transfer Learning Experiments section to explicitly confirm this and added a table listing the shared hyperparameters (optimizer, learning-rate schedule, layer-freezing strategy, augmentation pipeline, batch size, and early-stopping rule). These details were already described in the Methods but are now highlighted for clarity. revision: yes

  2. Referee: [Results and evaluation] Results and evaluation sections: The abstract asserts outperformance on the new dataset and in TL experiments but supplies no concrete metrics, baselines, error bars, statistical tests, or validation protocols. This prevents independent verification of the central claims and undermines the data-efficiency and robustness assertions.

    Authors: The Results and Evaluation sections contain tables with accuracy, F1, and other metrics comparing our model to baselines on the constructed dataset, along with learning curves and metrics for the transfer learning experiments on the second dataset. Validation used 5-fold cross-validation. To address the concern about the abstract and ease of verification, we have revised the abstract to include key quantitative results and added explicit mentions of error bars (from repeated runs) and statistical comparisons in the evaluation section. revision: yes

Circularity Check

0 steps flagged

No circularity; claims rest on empirical dataset construction and benchmarking

full rationale

The paper constructs a new dataset from public sources plus an augmentation study, trains DenseNet201 on it, and reports outperformance versus ImageNet baselines on both the new dataset and downstream TL tasks on a second dataset. No equations, self-citations, or uniqueness theorems are invoked to derive the result; performance is measured directly via training and evaluation. This is self-contained empirical work with no reduction of outputs to inputs by definition or construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper rests on standard computer-vision assumptions about CNN feature extraction and transfer learning benefits; no free parameters, new entities, or ad-hoc axioms are explicitly introduced in the abstract.

axioms (2)
  • domain assumption Convolutional neural networks can automatically extract useful features from leaf images for disease classification
    Invoked implicitly when stating that CNN models are efficient for this task without manual feature engineering.
  • domain assumption Transfer learning from a domain-specific pre-trained model yields faster, more stable training with less data than a general model
    Central to the claim that the new base model overcomes field issues; treated as established rather than derived.

pith-pipeline@v0.9.0 · 5594 in / 1356 out tokens · 64722 ms · 2026-05-09T14:48:26.398172+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

204 extracted references · 90 canonical work pages · 4 internal anchors

  1. [1]

    The State of Food Security and Nutrition in the World (SOFI)

    FAO, IFAD, UNICEF, WFP and WHO, The State of F ood Security and Nutrition in the World 2024 – Financing to end hunger , food insecurity and malnutrition in all its forms, ser. The State of Food Security and Nutrition in the World (SOFI). Rome, Italy: FAO, IFAD, UNICEF, WFP , WHO, 2024, no. 2024. doi: 10.4060/cd1254en

    work page doi:10.4060/cd1254en 2024

  2. [2]

    FAO Statistical Y earbook – World Food and Agriculture

    FAO, World F ood and Agriculture – Statistical Yearbook 2023 , ser. FAO Statistical Y earbook – World Food and Agriculture. Rome, Italy: FAO, 2023, no. 2023. doi: 10.4060/cc8166en

    work page doi:10.4060/cc8166en 2023

  3. [3]

    Rome, Italy: FAO, 2017

    ——, The Future of F ood and Agriculture: Trends and Challenges , 1st ed. Rome, Italy: FAO, 2017. [Online]. Available: https://openknowledge.fao.org/handle/20.500. 14283/i6583e

    2017

  4. [4]

    The persis- tent threat of emerging plant disease pandemics to global food security,

    J. B. Ristaino, P . K. Anderson, D. P . Bebber, K. A. Brauman, N. J. Cunniffe, N. V . Fedoroff, C. Finegold, K. A. Garrett, C. A. Gilligan, C. M. Jones et al. , “The persis- tent threat of emerging plant disease pandemics to global food security,” Proceedings of the National Academy of Sciences , vol. 118, no. 23, p. e2022239118, 2021. doi: 10.1073/pnas.2...

    work page doi:10.1073/pnas.2022239118 2021

  5. [5]

    The State of Food and Agriculture (SOFA)

    FAO, The State of F ood and Agriculture 2023 – Revealing the true cost of food to trans- form agrifood systems , ser. The State of Food and Agriculture (SOFA). Rome, Italy: FAO, 2023, no. 2023. doi: 10.4060/cc7724en

    work page doi:10.4060/cc7724en 2023

  6. [6]

    D. C. Clay, C. G. Carlson, and K. Dalsted, iGrow Wheat: Best Management Practices for Wheat Production . Brookings, S.D.: South Dakota State University, College of Agriculture and Biological Sciences. SDSU Extension, 2012, agronomy, 102 Horticulture, and Plant Science Books. 1., ISBN: 978-0-9856309-0-4. [Online]. Available: https://openprairie.sdstate.edu...

    2012

  7. [7]

    R. O. for Europe, Monitoring and Surveillance of Cereals Pests, Diseases and Weeds, 1st ed. Sub-regional Of fi ce of FAO for Central Asia: FAO, 2012, fAO Job Number: AQ295E. [Online]. Available: https://openknowledge.fao.org/handle/20.500. 14283/aq295e

    2012

  8. [8]

    Koyshibayev and H

    M. Koyshibayev and H. Muminjanov, GUIDELINES FOR MONITORING DISEASES, PESTS AND WEEDS IN CEREAL CROPS , 1st ed. Rome, Italy: FAO, 2016. [Online]. Available: https://openknowledge.fao.org/handle/20.500.14283/i5550e

    2016

  9. [9]

    Systematic review of deep learning techniques in plant disease detection,

    M. Nagaraju and P . Chawla, “Systematic review of deep learning techniques in plant disease detection,” International journal of system assurance engineering and manage- ment, vol. 11, no. 3, pp. 547–560, 2020. doi: 10.1007/s13198-020-00972-1

    work page doi:10.1007/s13198-020-00972-1 2020

  10. [10]

    Evaluation of performance, energy, and computation costs of quantum-attack resilient encryption algorithms for embedded de- vices,

    L. Li, S. Zhang, and B. Wang, “Plant disease detection and classi fi cation by deep learning—a review,” IEEE Access , vol. 9, pp. 56 683–56 698, 2021. doi: 10.1109/AC- CESS.2021.3069646

    work page doi:10.1109/ac- 2021

  11. [11]

    Review on convolutional neural network (cnn) applied to plant leaf disease classi fi cation,

    J. Lu, L. Tan, and H. Jiang, “Review on convolutional neural network (cnn) applied to plant leaf disease classi fi cation,” Agriculture, vol. 11, no. 8, p. 707, 2021. doi: 10.3390/agriculture11080707

    work page doi:10.3390/agriculture11080707 2021

  12. [12]

    Leaf disease detection using ma- chine learning and deep learning: Review and challenges,

    C. Sarkar, D. Gupta, U. Gupta, and B. B. Hazarika, “Leaf disease detection using ma- chine learning and deep learning: Review and challenges,” Applied Soft Computing, vol. 145, p. 110534, 2023. doi: 10.1016/j.asoc.2023.110534

    work page doi:10.1016/j.asoc.2023.110534 2023

  13. [13]

    Plant disease detection and classification techniques: a comparative study of the performances,

    W. B. Demilie, “Plant disease detection and classi fi cation techniques: a comparative study of the performances,” Journal of Big Data , vol. 11, no. 1, p. 5, 2024. doi: 10.1186/s40537-023-00863-9

    work page doi:10.1186/s40537-023-00863-9 2024

  14. [14]

    Embrac- ing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning,

    M. Xu, H. Kim, J. Y ang, A. Fuentes, Y . Meng, S. Y oon, T. Kim, and D. S. Park, “Embrac- ing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning,” Frontiers in Plant Science, vol. 14, p. 1225409, 2023. doi: 10.3389/fpls.2023.1225409 . 103

    work page doi:10.3389/fpls.2023.1225409 2023

  15. [15]

    Rady Plant Diseases Image- Text Pairs Default license (Hug- gingFace) No explicit license declared on the HuggingFace dataset card; contact author before redistribution

    D. Singh, N. Jain, P . Jain, P . Kayal, S. Kumawat, and N. Batra, “Plantdoc: A dataset for visual plant disease detection,” in Proceedings of the 7th ACM IKDD CoDS and 25th COMAD. New Y ork, NY , USA: Association for Computing Machinery, 2020, pp. 249–253. doi: 10.1145/3371158.3371196

    work page doi:10.1145/3371158.3371196 2020

  16. [16]

    Foundation models de fi ning a new era in vision: a survey and out- look,

    M. Awais, M. Naseer, S. Khan, R. M. Anwer, H. Cholakkal, M. Shah, M.-H. Y ang, and F. S. Khan, “Foundation models de fi ning a new era in vision: a survey and out- look,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2025. doi: 10.1109/TPAMI.2024.3506283

    work page doi:10.1109/tpami.2024.3506283 2025

  17. [17]

    A., Siebert, X., Cornu, O., and Vleeschouwer, C

    F. Chen, M. V . Giuffrida, and S. A. Tsaftaris, “Adapting vision foundation models for plant phenotyping,” in Proceedings of the IEEE/CVF International Conference on Com- puter Vision, 2023, pp. 604–613. doi: 10.1109/ICCVW60793.2023.00067

    work page doi:10.1109/iccvw60793.2023.00067 2023

  18. [18]

    A systematic review of the current state of transfer learning accelerated cnn-based plant leaf disease classi fi cation,

    D. J. Richter, M. I. Bappi, S. S. Kolekar, and K. Kim, “A systematic review of the current state of transfer learning accelerated cnn-based plant leaf disease classi fi cation,” 2025. doi: 10.1109/ACCESS.2025.3584404 ., in press

    work page doi:10.1109/access.2025.3584404 2025

  19. [19]

    and Hughes, David P

    S. P . Mohanty, D. P . Hughes, and M. Salath ´e, “Using deep learning for image- based plant disease detection,” Frontiers in plant science , vol. 7, p. 1419, 2016. doi: 10.3389/fpls.2016.01419

    work page doi:10.3389/fpls.2016.01419 2016

  20. [20]

    Image-based crop disease detection with federated learning,

    D. Mamba Kabala, A. Ha fi ane, L. Bobelin, and R. Canals, “Image-based crop disease detection with federated learning,” Scientifi c Reports, vol. 13, no. 1, p. 19220, 2023

    2023

  21. [21]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  22. [22]

    MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

    A. G. Howard, M. Zhu, B. Chen, D. Kalenichenko, W. Wang, T. Weyand, M. Andreetto, and H. Adam, “Mobilenets: Ef fi cient convolutional neural net- works for mobile vision applications,” arXiv preprint arXiv:1704.04861 , 2017. doi: 10.48550/arXiv.1704.04861 . 104

    work page internal anchor Pith review doi:10.48550/arxiv.1704.04861 2017

  23. [23]

    Transforming agriculture: Federated learning cnns for wheat disease severity assessment,

    S. Mehta, V . Kukreja, and A. Gupta, “Transforming agriculture: Federated learning cnns for wheat disease severity assessment,” in 2023 8th International Conference on Com- munication and Electronics Systems (ICCES) . IEEE, 2023, pp. 792–797

    2023

  24. [24]

    A novel deep learning model for detection of severity level of the disease in citrus fruits,

    P . Dhiman, V . Kukreja, P . Manoharan, A. Kaur, M. Kamruzzaman, I. B. Dhaou, and C. Iwendi, “A novel deep learning model for detection of severity level of the disease in citrus fruits,” Electronics, vol. 11, no. 3, p. 495, 2022

    2022

  25. [25]

    Federated learning-based uavs for the diagnosis of plant diseases,

    F. S. Khan, S. Khan, M. N. H. Mohd, A. Waseem, M. N. A. Khan, S. Ali, and R. Ahmed, “Federated learning-based uavs for the diagnosis of plant diseases,” in 2022 Interna- tional Conference on Engineering and Emerging Technologies (ICEET) . IEEE, 2022, pp. 1–6

    2022

  26. [26]

    Federated learning: Crop classi fi cation in a smart farm decentralised network,

    G. Idoje, T. Dagiuklas, and M. Iqbal, “Federated learning: Crop classi fi cation in a smart farm decentralised network,” Smart Agricultural Technology, vol. 5, p. 100277, 2023

    2023

  27. [27]

    Tomato plant disease detection and classi fi cation using convolutional neural network architectures technologies,

    D. R. Hammou and M. Boubaker, “Tomato plant disease detection and classi fi cation using convolutional neural network architectures technologies,” in Networking, Intelli- gent Systems and Security: Proceedings of NISS 2021 . Springer, 2022, pp. 33–44. doi: 10.1007/978-981-16-3637-0 3

    work page doi:10.1007/978-981-16-3637-0 2021

  28. [28]

    Zenseact open dataset: A large-scale and diverse multimodal dataset for autonomous driving

    B. Tej, F. Nasri, and A. Mtibaa, “Detection of pepper and tomato leaf diseases us- ing deep learning techniques,” in 2022 5th international conference on advanced systems and emergent technologies (IC ASET). IEEE, 2022, pp. 149–154. doi: 10.1109/IC ASET53395.2022.9765923

    work page doi:10.1109/ic 2022

  29. [29]

    Grapevine leaf disease identi fi cation using transfer learn- ing,

    R. Nagi and S. S. Tripathy, “Grapevine leaf disease identi fi cation using transfer learn- ing,” in 2021 IEEE International Women in Engineering (WIE) Conference on Elec- trical and Computer Engineering (WIECON-ECE) . IEEE, 2021, pp. 43–46. doi: 10.1109/WIECON-ECE54711.2021.9829703

    work page doi:10.1109/wiecon-ece54711.2021.9829703 2021

  30. [30]

    Apple leaf disease identi fi cation with a small and imbal- anced dataset based on lightweight convolutional networks,

    L. Li, S. Zhang, and B. Wang, “Apple leaf disease identi fi cation with a small and imbal- anced dataset based on lightweight convolutional networks,” Sensors, vol. 22, no. 1, p. 173, 2021. doi: 10.3390/s22010173 . 105

    work page doi:10.3390/s22010173 2021

  31. [31]

    Detection and classi fi cation of chilli leaf disease using a squeeze-and-excitation-based cnn model,

    B. N. Naik, R. Malmathanraj, and P . Palanisamy, “Detection and classi fi cation of chilli leaf disease using a squeeze-and-excitation-based cnn model,” Ecological Informatics, vol. 69, p. 101663, 2022. doi: 10.1016/j.ecoinf.2022.101663

    work page doi:10.1016/j.ecoinf.2022.101663 2022

  32. [32]

    Comparison of various deep convolutional neu- ral network models to discriminate apple leaf diseases using transfer learning,

    P . Pradhan, B. Kumar, and S. Mohan, “Comparison of various deep convolutional neu- ral network models to discriminate apple leaf diseases using transfer learning,” Jour- nal of Plant Diseases and Protection , vol. 129, no. 6, pp. 1461–1473, 2022. doi: 10.1007/s41348-022-00660-1

    work page doi:10.1007/s41348-022-00660-1 2022

  33. [33]

    Agroaid: A mobile app system for visual classi fi cation of plant species and diseases using deep learning and tensorfl ow lite,

    M. Reda, R. Suwwan, S. Alkafri, Y . Rashed, and T. Shanableh, “Agroaid: A mobile app system for visual classi fi cation of plant species and diseases using deep learning and tensorfl ow lite,” in Informatics, vol. 9. MDPI, 2022, p. 55. doi: 10.3390/informat- ics9030055

    work page doi:10.3390/informat- 2022

  34. [34]

    Adaptive thresholding of cnn features for maize leaf disease classi fi ca- tion and severity estimation,

    H. D. Mafukidze, G. Owomugisha, D. Otim, A. Nechibvute, C. Nyamhere, and F. Mazunga, “Adaptive thresholding of cnn features for maize leaf disease classi fi ca- tion and severity estimation,” Applied Sciences, vol. 12, no. 17, p. 8412, 2022

    2022

  35. [35]

    Deep learn- ing utilization in agriculture: Detection of rice plant diseases using an improved cnn model,

    G. Latif, S. E. Abdelhamid, R. E. Mallouhy, J. Alghazo, and Z. A. Kazimi, “Deep learn- ing utilization in agriculture: Detection of rice plant diseases using an improved cnn model,” Plants, vol. 11, no. 17, p. 2230, 2022. doi: 10.3390/plants11172230

    work page doi:10.3390/plants11172230 2022

  36. [36]

    Improving plant disease classi fi cation by adaptive minimal ensembling,

    A. Bruno, D. Moroni, R. Dainelli, L. Rocchi, S. Morelli, E. Ferrari, P . Toscano, and M. Martinelli, “Improving plant disease classi fi cation by adaptive minimal ensembling,” Frontiers in Arti fi cial Intelligence , vol. 5, p. 868926, 2022. doi: 10.3389/frai.2022.868926

    work page doi:10.3389/frai.2022.868926 2022

  37. [37]

    Deep learning-based leaf disease detection in crops using images for agricultural applications,

    J. Eunice, D. E. Popescu, M. K. Chowdary, and J. Hemanth, “Deep learning-based leaf disease detection in crops using images for agricultural applications,”Agronomy, vol. 12, no. 10, p. 2395, 2022. doi: 10.3390/agronomy12102395

    work page doi:10.3390/agronomy12102395 2022

  38. [38]

    Ef fi cient attention-based cnn network (eanet) for multi- class maize crop disease classi fi cation,

    S. Albahli and M. Masood, “Ef fi cient attention-based cnn network (eanet) for multi- class maize crop disease classi fi cation,” Frontiers in Plant Science, vol. 13, p. 1003152,

  39. [39]

    doi: 10.3389/fpls.2022.1003152 . 106

    work page doi:10.3389/fpls.2022.1003152 2022

  40. [40]

    Deep learning ensemble-based auto- mated and high-performing recognition of coffee leaf disease,

    D. Novtahaning, H. A. Shah, and J.-M. Kang, “Deep learning ensemble-based auto- mated and high-performing recognition of coffee leaf disease,” Agriculture, vol. 12, no. 11, p. 1909, 2022. doi: 10.3390/agriculture12111909

    work page doi:10.3390/agriculture12111909 1909

  41. [41]

    Improving deep learning-based plant disease classifi cation with attention mechanism,

    P . Alirezazadeh, M. Schirrmann, and F. Stolzenburg, “Improving deep learning-based plant disease classifi cation with attention mechanism,” Gesunde Pfl anzen, vol. 75, no. 1, pp. 49–59, 2023. doi: 10.1007/s10343-022-00796-y

    work page doi:10.1007/s10343-022-00796-y 2023

  42. [42]

    Logprompt: A log-based anomaly detection framework using prompts

    B. A. Fulle, C. Ma, X. Shi, W. Zhu, Z. Y emataw, and E. Assefa, “Ef fi cient early warning system in identifying enset bacterial wilt disease using transfer learning,” in 2023 In- ternational Joint Conference on Neural Networks (IJCNN) . IEEE, 2023, pp. 1–8. doi: 10.1109/IJCNN54540.2023.10191617

    work page doi:10.1109/ijcnn54540.2023.10191617 2023

  43. [43]

    Classi fi cation of rice leaf diseases using cnn-based pre-trained models and transfer learning,

    M. Mavaddat, M. Naderan, and S. E. Alavi, “Classi fi cation of rice leaf diseases using cnn-based pre-trained models and transfer learning,” in 2023 6th International Confer- ence on Pattern Recognition and Image Analysis (IPRIA) . IEEE, 2023, pp. 1–6. doi: 10.1109/IPRIA59240.2023.10147178

    work page doi:10.1109/ipria59240.2023.10147178 2023

  44. [44]

    In: Computer Vision – ECCV 2022 Workshops, pp

    E. Paiva-Peredo, “Deep learning for the classi fi cation of cassava leaf diseases in unbal- anced fi eld data set,” in International Conference on Advanced Network Technologies and Intelligent Computing . Springer, 2022, pp. 101–114. doi: 10.1007/978-3-031- 28183-9 8

    work page doi:10.1007/978-3-031- 2022

  45. [45]

    A precise model for accurate rice disease diagnosis: a trans- fer learning approach,

    D. Mohapatra and N. Das, “A precise model for accurate rice disease diagnosis: a trans- fer learning approach,” Proceedings of the Indian National Science Academy , vol. 89, no. 1, pp. 162–171, 2023. doi: 10.1007/s43538-022-00149-3

    work page doi:10.1007/s43538-022-00149-3 2023

  46. [46]

    An effective feature extraction method for olive peacock eye leaf disease classi fi ca- tion,

    A. Diker, A. Elen, C. K ¨ozkurt, S. K ılıc ¸arslan, E. D¨onmez, K. Arslan, and E. C. Kuran, “An effective feature extraction method for olive peacock eye leaf disease classi fi ca- tion,” European F ood Research and Technology, vol. 250, no. 1, pp. 287–299, 2024. doi: 10.1007/s00217-023-04386-8

    work page doi:10.1007/s00217-023-04386-8 2024

  47. [47]

    Modifi ed transfer learning frameworks to iden- tify potato leaf diseases,

    M. G. Lanjewar, P . Morajkar, and P . P , “Modifi ed transfer learning frameworks to iden- tify potato leaf diseases,”Multimedia Tools and Applications, vol. 83, no. 17, pp. 50 401– 50 423, 2024. doi: 10.1007/s11042-023-17610-0 . 107

    work page doi:10.1007/s11042-023-17610-0 2024

  48. [48]

    High-performance deep transfer learning model with batch normalization based on multiscale feature fusion for tomato plant disease identi fi cation and categorization,

    R. Ramya and P . Kumar, “High-performance deep transfer learning model with batch normalization based on multiscale feature fusion for tomato plant disease identi fi cation and categorization,” Environmental Research Communications, vol. 5, no. 12, p. 125015,

  49. [49]

    doi: 10.1088/2515-7620/ace594

    work page doi:10.1088/2515-7620/ace594

  50. [50]

    Effi cientrmt-net—an ef fi cient resnet-50 and vision transformers approach for clas- sifying potato plant leaf diseases,

    K. Shaheed, I. Qureshi, F. Abbas, S. Jabbar, Q. Abbas, H. Ahmad, and M. Z. Sajid, “Effi cientrmt-net—an ef fi cient resnet-50 and vision transformers approach for clas- sifying potato plant leaf diseases,” Sensors, vol. 23, no. 23, p. 9516, 2023. doi: 10.3390/s23239516

    work page doi:10.3390/s23239516 2023

  51. [51]

    Transfer learning-based deep learning model for corn leaf disease classi fi cation,

    J. An, N. Zhang, and W. H. Mahmoud, “Transfer learning-based deep learning model for corn leaf disease classi fi cation,” in International Symposium on Neural Networks . Springer, 2024, pp. 163–173. doi: 10.1007/978-981-97-4399-5 16

    work page doi:10.1007/978-981-97-4399-5 2024

  52. [52]

    Mulberry leaf disease detection using cnn-based smart android application,

    A. Salam, M. Naznine, N. Jahan, E. Nahid, M. Nahiduzzaman, and M. E. Chowdhury, “Mulberry leaf disease detection using cnn-based smart android application,” IEEE Ac- cess, 2024. doi: 10.1109/ACCESS.2024.3407153

    work page doi:10.1109/access.2024.3407153 2024

  53. [53]

    Ha and J

    R. K. Bania, N. A. Talukdar, and D. J. Bora, “Cutting-edge hybrid deep transfer learning approach for potato leaf disease classi fi cation,” International Journal of Agri- cultural and Natural Sciences , vol. 17, no. 1, pp. 111–123, 2024. doi: 10.5281/zen- odo.10727496

    work page doi:10.5281/zen- 2024

  54. [54]

    Application of improved lightweight network and cho- quet fuzzy ensemble technology for soybean disease identi fi cation,

    Y . Hang, X. Meng, and Q. Wu, “Application of improved lightweight network and cho- quet fuzzy ensemble technology for soybean disease identi fi cation,” IEEE Access, 2024. doi: 10.1109/ACCESS.2024.3365829

    work page doi:10.1109/access.2024.3365829 2024

  55. [55]

    Comprehensive mixed-based data aug- mentation for detection of rice leaf disease in the wild,

    A. L. A. Haikal, N. Y udistira, and A. Ridok, “Comprehensive mixed-based data aug- mentation for detection of rice leaf disease in the wild,” Crop Protection, p. 106816,

  56. [56]

    doi: 10.1016/j.cropro.2024.106816

    work page doi:10.1016/j.cropro.2024.106816 2024

  57. [57]

    Grape (vitis vinifera) leaf disease detection and classifi cation using deep learning techniques: A study on real-time grape leaf image dataset in india,

    S. K. Shah, V . Kumbhar, and T. Singh, “Grape (vitis vinifera) leaf disease detection and classifi cation using deep learning techniques: A study on real-time grape leaf image dataset in india,” International Journal of Engineering , vol. 37, no. 8, pp. 1522–1533,

  58. [58]

    doi: 10.5829/ije.2024.37.08b.06 . 108

    work page doi:10.5829/ije.2024.37.08b.06 2024

  59. [59]

    Densenet201plus: Cost-effective transfer-learning architecture for rapid leaf dis- ease identi fi cation with attention mechanisms,

    M. K. A. Mazumder, M. M. Kabir, A. Rahman, M. Abdullah-Al-Jubair, and M. Mridha, “Densenet201plus: Cost-effective transfer-learning architecture for rapid leaf dis- ease identi fi cation with attention mechanisms,” Heliyon, vol. 10, no. 15, 2024. doi: 10.1016/j.heliyon.2024.e35625

    work page doi:10.1016/j.heliyon.2024.e35625 2024

  60. [60]

    Identi fi cation of chili plant diseases based on leaves using hyperparameter optimization architecture convolutional neural network

    S. Winiarti, A. Pujiyanta et al. , “Identi fi cation of chili plant diseases based on leaves using hyperparameter optimization architecture convolutional neural network.” Interna- tional Journal of Advanced Computer Science & Applications , vol. 15, no. 11, 2024. doi: 10.14569/IJACSA.2024.0151185

    work page doi:10.14569/ijacsa.2024.0151185 2024

  61. [61]

    An enhanced lightweight t-net architecture based on convolutional neural network (cnn) for tomato plant leaf disease classifi cation,

    A. Batool, J. Kim, S.-J. Lee, J.-H. Y ang, and Y .-C. Byun, “An enhanced lightweight t-net architecture based on convolutional neural network (cnn) for tomato plant leaf disease classifi cation,” PeerJ Computer Science , vol. 10, p. e2495, 2024. doi: 10.7717/peerj- cs.2495

    work page doi:10.7717/peerj- 2024

  62. [62]

    ImageNet:

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in 2009 IEEE conference on computer vision and pattern recognition. Ieee, 2009, pp. 248–255. doi: 10.1109/CVPR.2009.5206848

    work page doi:10.1109/cvpr.2009.5206848 2009

  63. [63]

    J. E. Bidlack and S. Jansky, Stern’s introductory plant biology , 12th ed. New Y ork, USA: McGraw-Hill New Y ork, 2010, ISBN-10: 0073040525, ISBN-13: 978- 0073040523

    2010

  64. [64]

    Plant disease basics and diagnosis,

    Penn State Extension, “Plant disease basics and diagnosis,” 2012, accessed: 2024-11-11. [Online]. Available: https://extension.psu.edu/plant-disease-basics-and-diagnosis

    2012

  65. [65]

    A survey on the plant leaf disease detection techniques,

    A. Patel and B. Joshi, “A survey on the plant leaf disease detection techniques,” Inter- national Journal of Advanced Research in Computer and Communication Engineering , vol. 6, no. 1, pp. 229–231, 2017. doi: 10.17148/IJARCCE.2017.6143

    work page doi:10.17148/ijarcce.2017.6143 2017

  66. [66]

    From lab to fi eld: An empirical study on the generalization of convolutional neural networks towards crop disease detection,

    F. A. Guth, S. Ward, and K. McDonnell, “From lab to fi eld: An empirical study on the generalization of convolutional neural networks towards crop disease detection,” Euro- pean Journal of Engineering and Technology Research , vol. 8, no. 2, pp. 33–40, 2023. doi: 10.24018/ejeng.2023.8.2.2773 . 109

    work page doi:10.24018/ejeng.2023.8.2.2773 2023

  67. [67]

    Toward generalization of deep learning- based plant disease identi fi cation under controlled and fi eld conditions,

    A. Ahmad, A. El Gamal, and D. Saraswat, “Toward generalization of deep learning- based plant disease identi fi cation under controlled and fi eld conditions,” IEEE Access , vol. 11, pp. 9042–9057, 2023. doi: 10.1109/ACCESS.2023.3240100

    work page doi:10.1109/access.2023.3240100 2023

  68. [68]

    Deep learning models for plant disease detection and diagnosis,

    K. P . Ferentinos, “Deep learning models for plant disease detection and diagno- sis,” Computers and electronics in agriculture , vol. 145, pp. 311–318, 2018. doi: 10.1016/j.compag.2018.01.009

    work page doi:10.1016/j.compag.2018.01.009 2018

  69. [69]

    A novel dataset of potato leaf disease in uncontrolled environment,

    N. H. Shabrina, S. Indarti, R. Maharani, D. A. Kristiyanti, N. Prastomo et al., “A novel dataset of potato leaf disease in uncontrolled environment,” Data in Brief , vol. 52, p. 109955, 2024. doi: 10.1016/j.dib.2023.109955

    work page doi:10.1016/j.dib.2023.109955 2024

  70. [70]

    An open access repository of images on plant health to enable the development of mobile disease diagnostics,

    D. P . Hughes and M. Salathe, “An open access repository of images on plant health to enable the development of mobile disease diagnostics,” arXiv e-prints, pp. arXiv–1511,

  71. [71]

    doi: 10.48550/arXiv.1511.08060

    work page doi:10.48550/arxiv.1511.08060

  72. [72]

    Plantvillage dataset,

    S. P . Mohanty, “Plantvillage dataset,” https://github.com/spMohanty/ PlantVillage-Dataset, 2016, accessed: 2024-10-21

    2016

  73. [73]

    Plant disease detection with deep learning and feature extraction using plant village,

    F. Mohameth, C. Bingcai, and K. A. Sada, “Plant disease detection with deep learning and feature extraction using plant village,” Journal of Computer and Communications , vol. 8, no. 6, pp. 10–22, 2020

    2020

  74. [74]

    A novel deep learning method for detection and classi fi cation of plant diseases,

    W. Albattah, M. Nawaz, A. Javed, M. Masood, and S. Albahli, “A novel deep learning method for detection and classi fi cation of plant diseases,” Complex & Intelligent Sys- tems, pp. 1–18, 2022

    2022

  75. [75]

    “tomato-village

    M. Gehlot, R. K. Saxena, and G. C. Gandhi, ““tomato-village”: a dataset for end-to- end tomato disease detection in a real-world environment,” Multimedia Systems, vol. 29, no. 6, pp. 3305–3328, 2023. doi: 10.1007/s00530-023-01158-y

    work page doi:10.1007/s00530-023-01158-y 2023

  76. [76]

    Tomato-village dataset,

    M. Gehlot, “Tomato-village dataset,” https://www.kaggle.com/datasets/mamtag/ tomato-village, 2023, accessed: 2024-10-21

    2023

  77. [77]

    Tomato leaf disease classi fi cation by combining effi cientnetv2 and a swin transformer,

    Y . Sun, L. Ning, B. Zhao, and J. Y an, “Tomato leaf disease classi fi cation by combining effi cientnetv2 and a swin transformer,” Applied Sciences, vol. 14, no. 17, p. 7472, 2024. 110

    2024

  78. [78]

    Potato disease leaf dataset(pld),

    M. R. Saeed and D. J. Rashid, “Potato disease leaf dataset(pld),” 2021. doi: 10.34740/KAGGLE/DS/1562973 . [Online]. Available: https://www.kaggle.com/ds/ 1562973

    work page doi:10.34740/kaggle/ds/1562973 2021

  79. [79]

    Multi-level deep learning model for potato leaf disease recognition,

    J. Rashid, I. Khan, G. Ali, S. H. Almotiri, M. A. AlGhamdi, and K. Masood, “Multi-level deep learning model for potato leaf disease recognition,” Electronics, vol. 10, no. 17, p. 2064, 2021

    2064

  80. [80]

    tea sickness dataset,

    A. Kimutai, Gibson; F ¨orster, “tea sickness dataset,” 2022. doi: 10.17632/j32xdt2ff5.2 . [Online]. Available: https://data.mendeley.com/datasets/j32xdt2ff5/2

    work page doi:10.17632/j32xdt2ff5.2 2022

Showing first 80 references.