Comparison of Number and Size of Parameters for Retinal Disease Classification Models

A tabular dataset compares the performance and computational characteristics of a proposed lightweight CNN model against multiple state-of-the-art architectures. The data originates from experiments using 1,824 retinal fundus images from the Retinal Fundus Multi-Disease Image Dataset (RFMiD), split into three disease classes. The dataset was created by Usman Rafi and last updated on May 11, 2026.

Use Cases

• Benchmarking model efficiency based on reported parameter counts and inference times.
• Selecting architectures for deployment in low-resource clinical settings based on computational metrics.
• Analyzing the trade-off between model accuracy and resource consumption for retinal disease classification.
• Validating the scalability of lightweight CNNs for multi-class medical imaging problems.

Strengths

• Performance metrics for the proposed model are provided, including 90.33% testing accuracy and 4 ms inference time.
• The dataset includes results from a generalizability study using an external dataset, RFMiD 2.0.
• An ablation study was conducted to validate the primary design decisions of the proposed architecture.

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 at 9.5 KB, indicating limited scope.

Provenance

Source

figshare

Collection Method

Likely contains experimental results from training and evaluating multiple CNN architectures.

Freshness

Last updated 2026-05-11 17:32:47; freshness should be verified.