BAISen2: A Sentinel-2 image dataset for rip channels detection along the French Atlantic coast

Rip channels are deep incisions in surf-zone sandbars associated with rip currents, the leading cause of drowning worldwide. On meso- to macrotidal beaches, such bar–rip systems can emerge at low tide, as observed along the French Atlantic coast, where these rip channels are commonly called baïnes.Automated detection of these morphological features from remote sensing imagery could improve our understanding of their spatiotemporal evolution and help identify hazardous conditions along coastlines. Yet no open-access satellite-imagery dataset dedicated to rip-channel morphology has been published. Here we introduce the BAISen2 (Baïnes Sentinel-2 dataset), an object detection dataset built from Sentinel-2 multispectral imagery covering 350 km along the French Atlantic shoreline spanning 2015–2024. The dataset comprises 2,058 image patches, of which 1,850 contain at least one annotated rip channel, for a total of 12,504 bounding-box annotations. The dataset is released in YOLO and COCO formats. BAISen2 is intended to support the development of automated coastal monitoring tools and is publicly available.