Performance Metrics for Rice Leaf Disease Detection Models

Ummer Shakeel published a dataset on figshare in 2026 containing performance metrics for deep learning models detecting rice leaf diseases. The dataset likely contains comparative evaluation results for five transfer learning architectures, including InceptionV3, DenseNet201, and MobileNetV2, trained on a balanced set of 1914 image samples. The work focuses on Pakistan, a major rice producer, and uses post-hoc visualization techniques like GrabCut segmentation for interpretability.

Use Cases

• Benchmarking deep transfer learning models for image classification based on reported training, validation, and test accuracies.
• Comparing the performance of lightweight versus high-density neural network architectures for agricultural tasks.
• Analyzing the effectiveness of post-hoc visualization techniques like GrabCut segmentation for model interpretability in plant disease detection.

Strengths

• Provides specific accuracy metrics for five distinct deep learning models, such as InceptionV3's 98.43% test accuracy.
• Based on a balanced dataset of 1914 image samples, which suggests a structured foundation for model training.
• Includes results for models trained over 30 epochs, offering insight into training stability.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for statistical analysis.
• The 5.5 KB file size indicates the dataset contains only summary metrics, not the underlying image data.

Provenance

Source

figshare

Collection Method

Results from systematic processing of a publicly available rice leaf disease image dataset using Python with TensorFlow and GPU computation.

Time Range

null

Freshness

Last updated 2026-05 17:41:10; freshness should be verified.

Geography

The study context focuses on Pakistan, a major rice producer.