A Canada-Wide Benchmark Dataset for Mining Detection from Satellite Imagery

This dataset provides a Canada-wide benchmark dataset for mining detection from high-resolution RGB satellite imagery. The dataset contains 1,519 manually annotated image patches (750 × 750 pixels), including both pixel-level segmentation masks and object detection bounding boxes. The dataset supports research in deep learning-based mining detection, segmentation, and object detection tasks. The RGB imagery was manually captured from Google Earth between 14 January 2025 and 7 February 2025. Mining locations were identified using the FINEPRINT visualizer (https://www.fineprint.global/visualisations/viewer/), which was used solely as a geographic reference tool. All RGB images were obtained directly from Google Earth.