RFMiD: Class-wise Performance Metrics for a Retinal Disease CNN

Class-wise performance metrics for the LiteFeatNet convolutional neural network evaluated on retinal fundus images. The dataset contains results from experiments using 1,824 images across three disease classes from the Retinal Fundus Multi-Disease Image Dataset (RFMiD), split 60:20:20 for training, validation, and testing. It was authored by Usman Rafi and last updated on May 11, 2026.

Use Cases

• Benchmarking lightweight CNN architectures based on reported accuracy, precision, recall, and F1-scores.
• Analyzing model performance across specific retinal disease categories mentioned in the description.
• Evaluating computational efficiency for clinical deployment based on reported model size and inference time.
• Comparing feature extraction and refinement methods based on the described ablation study.

Strengths

• Metrics are derived from a model achieving 90.33% testing accuracy on a defined test set.
• Results include precision, recall, and F1-score, providing a multi-faceted performance view.
• Model evaluation includes inference time (4 ms per image) and size (19.87 MB), relevant for deployment.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (9.5 KB), indicating limited scope, likely containing only summary statistics.

Provenance

Source

figshare

Collection Method

Derived from experiments on the Retinal Fundus Multi-Disease Image Dataset (RFMiD).

Freshness

Last updated 2026-05-11 17:32:50; freshness should be verified.