Work · Agricultural vision
Rice leaf ensembles
- Demo ready
- Synthetic demonstration
- Related literature only
A feature baseline, a transfer model and a validation-selected ensemble on a small public leaf-disease benchmark, with duplicate checks and error inspection.
Problem
With 120 images, an ensemble can look better than its parts purely through split luck. The repository keeps the weights on validation data and reports how wide the uncertainty is.
Contribution
Relationship to research: Related literature only. Cites a rice-leaf ensemble paper as related literature only; not an implementation of it.
Account owner’s role: Commissioned, reviewed and published the implementation. The code was produced with AI-assisted generation on commission and reviewed before release; it is not the original code of any cited publication.
What was built
- Colour and texture feature baseline with logistic regression
- ResNet-18 transfer model
- Ensemble weights chosen on validation only
- Exact and near-duplicate checks before evaluation
Method
Stratified repeated seeded splits; augmentation after splitting, training only.
Data and access
UCI Rice Leaf Diseases (CC BY 4.0), obtained by the user. Acquisition groups are unknown, so grouped splits cannot be formed.
Evaluation protocol
Class-wise recall, macro F1, confusion matrix, bootstrap intervals over test images.
Results
No measured result is published for this project. Demonstration outputs from the repository’s own synthetic fixtures are labelled as such in the repository and are not presented here as results.
Limitations
- Repeated splits are not independent field cohorts.
- The dataset is small and its collection conditions undocumented.
Reproducibility
Offline synthetic smoke; real data by local path only.
Repository metadata: Python, MIT, last push 2026-09-13, 0 stars.