Image Augmentation Parameters for Retinal Fundus Multi-Disease Image Dataset

1824 retinal fundus images from three disease classes were used to train the LiteFeatNet model. The dataset, created by Usman Rafi and last updated in May 2026, contains parameters likely used for augmenting the image data. The model achieved a testing accuracy of 90.33% and an inference time of 4 milliseconds per image.

Use Cases

• Benchmarking lightweight CNN architectures based on the reported performance metrics.
• Replicating or extending the image augmentation pipeline described for the RFMiD dataset.
• Training efficient models for multi-disease classification from retinal fundus images.
• Conducting ablation studies on feature extraction and refinement modules for CNNs.

Strengths

• Model performance is quantified with specific metrics: 90.33% accuracy, 90.69% precision, 90.27% F1-score.
• Computational efficiency is documented with a model size of 19.87 MB and an inference time of 4 milliseconds per image.
• The study includes a generalizability test using an external dataset (RFMiD 2.0) and a scalability test with additional disease categories.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small at 5.5 KB, indicating it contains parameters, not the original image data.

Provenance

Source

figshare

Collection Method

Parameters derived from experiments using the Retinal Fundus Multi-Disease Image Dataset (RFMiD).

Freshness

Last updated 2026-05-11 17:32:44; freshness should be verified.