Côte d'Ivoire Artisanal and Small-Scale Mining (ASM) Tile Dataset for Cross-Country Deep Learning Transferability Research

Artisanal and small-scale mining (ASM) is one of the fastest-expanding forms of land-use change in West Africa, driving deforestation, water contamination, and governance challenges across the region. This dataset supports research into deep learning-based ASGM segmentation and, specifically, the cross-national transferability of segmentation models trained in one country and deployed in another. Data Description The dataset contains 3,302 image tiles and 3,302 paired binary mining masks covering ASGM sites in Côte d'Ivoire, derived from 2024 Sentinel-2 and Sentinel-1 acquisitions at 10 m spatial resolution. Of the 3,302 tiles, 608 are mine-positive (contain at least one ASGM-labelled pixel) and form the study subset used in the associated paper. All tiles are 128 × 128 pixel GeoTIFF. Each image tile is a composite of 12 bands that include: Multispectral data from Sentinel-2: B2, B3, B4, B5, B6, B7, B8, B8A, B11, B12 SAR backscatter from Sentinel-1: VV, VH (converted to dB scale) The dataset is structured as follows: IVC_tiles.zip └── data/ └── IVC_tiles/ ├── west_zone/ │ ├── IMAGE/ IMG_IVC_*.tif │ └── MASK/ MASK_IVC_*.tif ├── east_zone/ │ ├── IMAGE/ IMG_IVC_*.tif │ └── MASK/ MASK_IVC_*.tif └── central_zone/ ├── IMAGE/ IMG_IVC_*.tif └── MASK/ MASK_IVC_*.tif Class mapping used in the masks: Class Label value Description 1 0 Non-mining (background) 2 1 ASGM pixel Ghana Data This release covers Côte d'Ivoire only. The Ghana tiles used in the associated study come from SmallMinesDS (Ofori-Ampofo et al., IEEE GRSL 2025, DOI: 10.1109/LGRS.2025.3566356), available at huggingface.co. Cite SmallMinesDS separately if you use Ghana data. Citation If you use this dataset, please cite: License CC-BY-4.0.