A Sentinel-2 multi-label dataset for waterway, port, and railway infrastructure classification across Germany

This dataset contains 10,000 Sentinel-2 true-color image chips for multi-label classification of waterway, port, and railway infrastructure across Germany. Each image is annotated independently with three binary labels: waterway, port area, and railway infrastructure. These labels result in eight possible combinations, including images containing none of the three categories. Each image is provided as a 224 × 224 pixel RGB PNG file and represents a 512 m × 512 m geographic footprint. The dataset contains 8,000 positive and 2,000 no-label image chips. The final distribution comprises 2,138 waterway-only images, 24 port-area-only images, 2,139 railway-infrastructure-only images, 115 waterway-and-port-area images, 2,138 waterway-and-railway-infrastructure images, 618 port-area-and-railway-infrastructure images, 828 images containing all three labels, and 2,000 images containing none of the three dataset categories. The distribution reflects the limited availability and strong spatial co-occurrence of industrial port areas with waterways and railway infrastructure in the retained OpenStreetMap source data. Label annotations were derived from OpenStreetMap geometries and recomputed for the complete footprint of every image. The waterway category includes geometries tagged waterway=river, waterway=canal, waterway=riverbank, and natural=water combined with water=river or water=canal. The port-area category includes industrial port areas tagged landuse=industrial in combination with industrial=port. Marinas, areas tagged only with harbour=yes, and small piers or jetties without industrial port land use were excluded from the port-area definition. The railway-infrastructure category is based on geometries tagged railway=rail. Tramways, subways, narrow-gauge railways, and disused, abandoned, proposed, razed, or under-construction railway segments were excluded. Railway segments marked as tunnels were also excluded from positive label assignment. The three labels are not mutually exclusive. A single image may contain any combination of waterways, industrial port areas, and railway infrastructure and receives all applicable labels. Source candidates were drawn from retained OpenStreetMap waterways, industrial port areas, and railway lines. Candidate locations were spatially distributed across Germany, and labels were determined from the complete 512 m × 512 m image footprints using the specified visibility thresholds. Images without any of the three labels were generated from neighboring locations in the surroundings of retained transport networks. They were accepted only when their complete footprints did not overlap with any retained waterway, port-area, or railway-infrastructure geometry. The absence of all three labels therefore denotes the absence of the specific categories included in the dataset definition; it does not imply the absence of all water bodies, transportation features, or human-made infrastructure. Streams, ditches, drains, lakes, ponds, reservoirs, and coastal waters were excluded from the waterway definition. Residential waterways, small drainage features, marinas, piers, non-industrial harbors, tramways, subways, and other railway categories not included in the dataset definition may therefore still be visible in images without the corresponding labels. Image chips were generated from Sentinel-2 Level-2A imagery using bands B04, B03, and B02 as red, green, and blue channels. Imagery was retrieved through the Copernicus Data Space Ecosystem Sentinel Hub Process API for the period from July 6, 2025, to July 5, 2026. The collection was generated using a maximum cloud-cover threshold of 10 percent and least-cloud-cover mosaicking. The repository includes the complete image collection and a compact CSV table containing image filenames and the three binary labels. The dataset can be used for supervised multi-label image classification, benchmarking of computer vision models, representation learning, waterway and railway mapping, port and logistics infrastructure analysis, and remote-sensing research on intermodal transportation systems.