太阳能电池EL图像缺陷检测数据集（PVEL-AD）

由于数据集受到大量关注和引用，应多数人要求，现将test标签一并公开，望大家周知！ Due to the growing interest and widespread usage of this dataset, and in response to numerous requests from the community, we have decided to publicly release the test set annotations as well.

新闻

[2026-1-30]：数据集网盘下载地址【new】google drive：https://drive.google.com/file/d/1EtteKnLhSFQ3XMCRXt5wKY-lDkIP7299/view?usp=drive_link

[2023-12-19]：2周内回复！Reply within 2 weeks! I am busy with my graduation thesis, please understand!

[2023-03-06]：所有人须知: 我们没有提供测试标注，测试你的算法请到 https://www.kaggle.com/competitions/pvelad

[2023-03-06]：Notice to all: We do not provide test annotations, please go to the following URL for testing your algorithms https://www.kaggle.com/competitions/pvelad

[2022-04-13]：Box annotations for vertical_dislocation and horizontal_dislocation will be added into PVELAD dataset.

[2021-12-14]: Training data augmentation via horizontal_flipping.py. Evaluation: first, converting ground truth xml to txt by get_gt_txt.py; Second, appling AP50-5-95.py to evaluate the detection results.

[2021-11-23]: A kaggle competition platform is built, then you can submit you result in https://www.kaggle.com/c/pvelad, and evaluate your algorithm.

数据集申请网站: http://aihebut.com/col.jsp?id=118 or https://github.com/binyisu/PVEL-AD

2021年数据集获取说明:

我们构建了一个用于太阳能电池的光伏电致发光异常检测数据集（PVEL-AD），该数据集包含36,543张具有各种内部缺陷和复杂背景的近红外图像。此数据集包含1类无异常图像和具有12种不同类别异常的图像，例如裂纹（线状和星状）、栅线中断、黑心、粗线、划痕、碎片、边角缺陷、印刷错误、水平错位、垂直错位和短路缺陷。此外，还提供了针对12种缺陷类型的40358个真实边界框标注。这是一项长尾目标检测任务，对于智能制造具有挑战性和重要意义。

类别（12类）	trainval	test
finger	2958	22638
crack	1260	2797
black_core	1028	3877
thick_line	981	1585
horizontal_dislocation	798	1582
short_circuit	492	1215
vertical_dislocation	137	271
star_crack	135	83
printing_error	32	48
corner	9	12
fragment	7	5
scratch	5	3

PVELAD-2021数据集申请表可在此处获取。

所有研究人员需按照以下说明获取数据集：

• 下载并填写工业数据集申请表（必须手写签名并注明日期）。请使用机构邮箱。不允许使用Gmail、QQ邮箱等商业邮箱。

• 将 signed 工业数据集申请表发送至 subinyi@vip.qq.com

• 请注意，若您希望通过谷歌云盘下载，请将您的谷歌邮箱一并发送给我。

• 该数据集由河北工业大学和北京航空航天大学联合发布。

[1] Binyi Su, Zhong Zhou, Haiyong Chen, “PVEL-AD: A Large-Scale Open-World Dataset for Photovoltaic Cell Anomaly Detection,” IEEE Trans. Ind. Inform., DOI (identifier) :10.1109/TII.2022.3162846

[2] B. Su, H. Chen, Y. Zhu, W. Liu and K. Liu, ``Classification of Manufacturing Defects in Multicrystalline Solar Cells With Novel Feature Descriptor,'' IEEE Trans. Instrum. Meas., vol. 68, no. 12, pp. 4675--4688, Dec. 2019.

[3] B. Su, H. Chen, and P. Chen, ``Deep Learning-Based Solar-Cell Manufacturing Defect Detection With Complementary Attention Network,'' IEEE Trans. Ind. Inform., vol. 17, no. 6, pp. 4084--4095, Jun. 2021.

[4] B. Su, H. Chen, and Z. Zhou, ``BAF-Detector: An Efficient CNN-Based Detector for Photovoltaic Cell Defect Detection,'' IEEE Trans. Ind. Electron., vol. 69, no. 3, pp. 3161-3171, Mar. 2022.