UEX-PVSeg: A Benchmark for Photovoltaic Panel Segmentation in Remote Sensing

Dataset Description

UEX-PVSeg is a comprehensive benchmark for photovoltaic (PV) panel segmentation in remote sensing imagery. It was established to facilitate rigorous evaluation of computer vision models on PV interpretation tasks, particularly for semantic segmentation. The benchmark is a key contribution of the paper "PANEL: A Domain-Specific Vision-Language Model for Zero-Shot and Few-Shot Photovoltaic Tasks in Remote Sensing."

Dataset Summary

The benchmark is stratified into three distinct datasets based on spatial resolution bands, ensuring a diverse and challenging evaluation environment:

• PV-SegA (0.1-0.3m): High-resolution imagery focusing on distributed rooftop PV installations.
• PV-SegB (0.6-1.0m): Medium-resolution imagery covering both rooftop and small-scale utility PV systems.
• PV-SegC (10-20m): Low-resolution imagery targeting large-scale utility PV farms.

This stratification allows researchers to assess model performance across different scales and contexts, from dense urban environments to vast desert landscapes.

Dataset Structure

Subset	Resolution (m)	Crop Size (px)	Count (Positives)	Source Composition & References
PV-SegA	0.1, 0.15, 0.2, 0.3	400	48,962	BDAPPV [1], Heilbronn [2], PV 01/03 [3], Fresno, Modesto, Stockton [4]
PV-SegB	0.6, 0.8, 1.0	512	6,034	Global PV-2022 (0.6m) [5], PV 08 [3], Oxnard [4], AIR-PV [6]
PV-SegC	10, 20	256	507,456	China PV-2022 [7], Global PV-2022 (20m) [5]

Data Sources and References

The following datasets were integrated to form the UEX-PVSeg benchmark:

• [1] BDAPPV: G. Kasmi, Y.-M. Saint-Drenan, D. Trebosc, R. Jolivet, J. Leloux, B. Sarr, L. Dubus, A crowdsourced dataset of aerial images with annotated solar photovoltaic arrays and installation metadata, Scientific Data 10 (1) (2023) 59.
• [2] Heilbronn: P. Li, H. Zhang, Z. Guo, S. Lyu, J. Chen, W. Li, X. Song, R. Shibasaki, J. Yan, Understanding rooftop pv panel semantic segmentation of satellite and aerial images for better using machine learning, Advances in Applied Energy 4 (2021) 100057.
• [3] PV 01/03/08: H. Jiang, L. Yao, N. Lu, J. Qin, T. Liu, Y. Liu, C. Zhou, Multi-resolution dataset for photovoltaic panel segmentation from satellite and aerial imagery, Earth System Science Data Discussions 2021 (2021) 1–17.
• [4] USA (Fresno, etc.): K. Bradbury, R. Saboo, T. L. Johnson, J. M. Malof, A. Devarajan, W. Zhang, L. M. Collins, R. G. Newell, Distributed solar photovoltaic array location and extent dataset for remote sensing object identification, Scientific Data 3 (1) (2016) 1–9.
• [5] Global PV-2022: A. Li, L. Liu, S. Li, X. Cui, X. Chen, X. Cao, Global photovoltaic solar panel dataset from 2019 to 2022, Scientific Data 12 (1) (2025) 637.
• [6] AIR-PV: Z. Yan, P. Wang, F. Xu, X. Sun, W. Diao, Air-pv: A benchmark dataset for photovoltaic panel extraction in optical remote sensing imagery, Science China Information Sciences 66 (4) (2023) 140307.
• [7] China PV-2022: Y. Chen, J. Zhou, Y. Ge, J. Dong, Uncovering the rapid expansion of photovoltaic power plants in china from 2010 to 2022 using satellite data and deep learning, Remote Sensing of Environment 305 (2024) 114100.

Data Instances

Each instance in the dataset consists of an image and its corresponding segmentation mask.

{
  "image": <PIL.Image.Image image mode=RGB size=400x400>,
  "mask": <PIL.Image.Image image mode=L size=400x400>
}

Data Fields

• image: The input remote sensing image.
• mask: The ground truth binary segmentation mask, where PV panels are marked.

Additional Information

Licensing Information

The UEX-PVSeg benchmark is released under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license. Users are free to share and adapt the data for non-commercial purposes, provided they give appropriate credit.

Citation Information

If you use the UEX-PVSeg benchmark in your research, please cite the following paper:

@article{deng2026panel,
  title={PANEL: A Domain-Specific Vision-Language Model for Zero-Shot and Few-Shot Photovoltaic Tasks in Remote Sensing (Under Review)},
  author={Deng, Ruizhe and Guo, Zhiling and Zhang, Penglei and Li, Jiaze and Xu, Xin and Chen, Qi and Chen, Yuntian and Yan, Jinyue},
  journal={ISPRS Journal of Photogrammetry and Remote Sensing},
  year={2026},
  publisher={Elsevier}
}

Dataset Curators

This benchmark was curated by the International Centre of Urban Energy Nexux (UEX) at The Hong Kong Polytechnic University and Eastern Institute of Technology (Ningbo), as part of the PANEL project.

Related Models

The PANEL model, introduced in the same paper, was trained and evaluated using this benchmark. It is a domain-specific Vision-Language Model for PV tasks and is available at https://huggingface.co/UEXdo/PANEL.

Collection

This dataset is part of the UEX-RenewableEnergy Collection.