A Nationwide Multi-Resolution Dataset for Photovoltaic Arrays based on Satellite and Unmanned Aerial Vehicle Imagery

A large-scale nationwide photovoltaic array dataset derived from multi-source satellite and UAV imagery. The dataset comprises samples collected at multiple spatial resolutions ranging from 0.3 m to 2 m, specifically including high-resolution products from Gaofen-2 (0.8 m & 2 m), Jilin-1 (0.75 m & 2 m), and detailed aerial photography via UAVs (0.3 m). Distinct from localized benchmarks, this dataset covers heterogeneous landscapes across mainland China. To address scale variations effectively within uniform image tiles, samples are categorized based on their proportional coverage: tiles containing fewer than 4,000 PV pixels (approximately 1.5% of the total area) are classified as "Positive-small" to represent small-scale distributed arrays, while those exceeding this threshold are designated as "Positive-large," corresponding to utility-scale centralized plants. Finally, the dataset adopts a hard-negative enriched design to mitigate false-positive risks triggered by complex background interference.NOTE: Some large files are split into ~2 GB parts due to upload constraints. To reconstruct the original files, download all parts with the same prefix (e.g., Gaofen-2_2m.part3_rar.part1.rar, part2.rar, part3.rar) and use WinRAR to open part1.rar — it will automatically read the subsequent parts.