Model Accuracy Comparison for Rice Leaf Disease Detection

Pakistan is a major global rice producer. This dataset contains a comparative evaluation of five deep transfer learning architectures (InceptionV3, DenseNet201, ResNet152V2, EfficientNetV2L, MobileNetV2) trained on a balanced set of 1914 rice leaf images for disease detection. The dataset was created by Ummer Shakeel and last updated in May 2026.

Use Cases

• Benchmarking deep learning model performance for image classification based on reported training, validation, and test accuracies.
• Evaluating the trade-off between model complexity and accuracy for lightweight deployment based on the comparison of architectures like MobileNetV2.
• Developing interpretable AI systems for agriculture by applying post-hoc visualization techniques like GrabCut segmentation and contour detection mentioned in the description.

Strengths

• Includes performance metrics for five distinct deep learning models, with specific accuracy percentages reported for training, validation, and testing phases.
• Based on a balanced dataset of 1914 image samples, which suggests a structured approach to model training.
• Uses a publicly available source dataset from Zenodo, supporting research reproducibility.

Limitations

• The dataset is very small at 5.5 KB, indicating it likely contains only summary results or metadata, not the underlying image data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for certain analyses.

Provenance

Source

figshare

Collection Method

Models were trained using Python with TensorFlow and GPU computation on a rice leaf image dataset from Zenodo.

Time Range

The study involved training models over 30 epochs; specific data collection dates are not provided.

Freshness

Last updated 2026-05-07 17:41:07; freshness should be verified.

Geography

The research context is Pakistan, a major rice producer, though the specific geographic origin of the image data is not stated.