RFMiD: Performance Metrics for a Lightweight Retinal Disease Detection Model

Usman Rafi published performance metrics for the LiteFeatNet model on May 11, 2026. The data includes testing accuracy, precision, recall, F1-score, and inference time results from experiments on the Retinal Fundus Multi-Disease Image Dataset (RFMiD). The model was evaluated on 1,824 images across three disease classes and compared against twelve state-of-the-art architectures.

Use Cases

• Benchmarking lightweight convolutional neural network performance based on reported accuracy and inference time.
• Comparing model efficiency for retinal disease classification based on the number of trainable parameters and computational resource requirements.
• Evaluating the scalability of integrated frameworks for multi-class medical imaging problems based on the described study with additional disease categories.
• Assessing model generalizability using external validation datasets based on the mentioned study with RFMiD 2.0.

Strengths

• Performance metrics are directly reported, including a testing accuracy of 90.33% and an inference time of 4 milliseconds per image.
• The dataset is small (9.5 KB), facilitating quick download and inspection.
• Experiments were conducted on a defined dataset of 1,824 images with a clear 60:20:20 train-validation-test split.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset contains only aggregated performance metrics, not the underlying image data or model parameters.

Provenance

Source

figshare

Collection Method

Likely contains experimental results from training and evaluating the LiteFeatNet convolutional neural network.

Freshness

Last updated 2026-05-11 17:32:49; freshness should be verified.