Baktiyar Ablaikhan
λ 443.9 nm

Deep-SAR Oil Spill Segmentation (Refined)

Satellite radar images of the sea with oil-spill masks corrected by hand — 38% of training masks and half of the validation masks.

Period
Feb 2026
Role
Curator
Status
Published dataset · CC BY 4.0
Six SAR patches from the dataset’s page — dark streaks are oil on the sea surface
of training masks corrected by hand
38%
of training masks corrected by hand
of validation masks refined
50%
of validation masks refined
SAR images, used in a peer-reviewed paper in Discover Computing (Springer Nature, 2026)
8,070
SAR images, used in a peer-reviewed paper in Discover Computing (Springer Nature, 2026)

The original Deep-SAR Oil Spill dataset (Q. Zhu, Y. Zhang, Z. Li and colleagues) pairs SAR images with binary oil/sea masks; some masks were wrong. I corrected them by hand and republished the set.

Used in

Jeena Joseph, “An edge-guided hybrid CNN–transformer framework for accurate marine oil spill segmentation in synthetic aperture radar images”, Discover Computing (Springer Nature), published online 17 August 2026. The paper trains and tests its model on the refined dataset’s 8,070 images (6,455 training, 1,615 validation) and cites it in its data-availability statement.