x-brick-kiln

This dataset contains 17,000 paired Sentinel-2 multispectral and Sentinel-1 SAR image patches (64×64 pixels) aimed at detecting brick kilns in Bangladesh. As the original dataset did not contain temporal extent information for the imagery, this version includes a refined temporal alignment process: For each Sentinel-2 image, all available cloud-free acquisitions between October 2018 and May 2019 were collected. The most temporally relevant date to the Sentinel-2 image was identified, and for that date, the corresponding Sentinel-1 SAR image was downloaded. This approach ensures a closer temporal match between the multispectral and SAR data. The dataset includes 13 spectral bands from Sentinel-2 (B1–B12) and two SAR polarizations (VV and VH) from Sentinel-1. Each image patch is labeled as either containing a brick kiln (label 1) or not (label 0), based on ground truth data from Lee et al. (2021) and refined using high-resolution imagery. Data is provided in HDF5 format, including image data, coordinates, tile locations, and binary labels. This dataset supports scalable monitoring of brick kiln locations, a significant contributor to pollution and carbon emissions in South Asia.