Plant Leaf Disease Detection Using Deep Learning: A Multi-Dataset Approach
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Agricultural productivity is increasingly threatened by plant diseases, which can spread rapidly and lead to significant crop losses if not identified early. Detecting plant diseases accurately in diverse and uncontrolled environments remains challenging, as most current detection methods rely heavily on lab-captured images that may not generalise well to real-world settings. This paper aims to develop models capable of accurately identifying plant diseases across diverse conditions, overcoming the limitations of existing methods. A combined dataset was utilised, incorporating the PlantDoc dataset with web-sourced images of plants from online platforms. State-of-the-art convolutional neural network (CNN) architectures, including EfficientNet-B0, EfficientNet-B3, ResNet50, and DenseNet201, were employed and fine-tuned for plant leaf disease classification. A key contribution of this work is the application of enhanced data augmentation techniques, such as adding Gaussian noise, to improve model generalisation. The results demonstrated varied performance across the datasets. When trained and tested on the PlantDoc dataset, EfficientNet-B3 achieved an accuracy of 73.31%. In cross-dataset evaluation, where the model was trained on PlantDoc and tested on a web-sourced dataset, EfficientNet-B3 reached 76.77% accuracy. The best performance was achieved with the combination of the PlanDoc and web-sourced datasets resulting in an accuracy of 80.19% indicating very good generalisation in diverse conditions. Class-wise F1-scores consistently exceeded 90% for diseases such as apple rust leaf and grape leaf across all models, demonstrating the effectiveness of this approach for plant disease detection.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.003 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it