<b>Gated Pathwise SSM2D</b>: <i>DRIVE128</i> and <i>Snake128</i> Segmentation Datasets

Chelvyn Immanuel published this 101.5 MB dataset on figshare in April 2026. It contains two preprocessed image datasets for curvilinear structure segmentation. The first is a processed version of the DRIVE retinal vessel dataset, and the second is a synthetically generated 'snake' dataset.

Use Cases

• Training segmentation models based on preprocessed 128x128 retinal image crops.
• Benchmarking model performance on synthetic curvilinear structures based on the generated snake dataset.
• Reproducing experiments from the associated paper on 2D State Space Models for segmentation.

Strengths

• Includes a well-known benchmark dataset (DRIVE) with a clear citation to the original 2004 paper.
• Data is preprocessed to a consistent 128x128 resolution, likely reducing initial setup effort.
• Provides a complementary synthetic dataset (Snake128) for controlled experimentation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The synthetic dataset may not fully capture the complexity of real-world medical images.

Provenance

Source

figshare, authored by Chelvyn Immanuel.

Collection Method

One dataset is a preprocessed version of the DRIVE benchmark; the other is synthetically generated for the study.

Time Range

Original DRIVE dataset from 2004; preprocessing and synthetic generation date unknown.

Freshness

Last updated 2026-04-04 11:03:15; freshness should be verified.

Geography

Original DRIVE dataset focuses on retinal images; geographic scope of the synthetic data is not specified.