Solar cell EL image defect detection dataset
News
[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.
Dataset application website: http://aihebut.com/col.jsp?id=118 or https://github.com/binyisu/PVEL-AD
2021 Dataset Access Instructions:
We build a Photovoltaic Electroluminescence Anomaly Detection dataset (PVEL-AD ) for solar cells, which contains 36,543 near-infrared images with various internal defects and heterogeneous backgrounds. This dataset contains 1 class of anomaly-free images and anomalous images with 12 different categories such as crack (line and star), finger interruption, black core, thick line, scratch, fragment, corner, printing_error, horizontal_dislocation, vertical_dislocation, and short_circuit defects. Moreover, 40358 ground truth bounding boxes are provided for 12 types of defects. This is a long-tail object detection task, which is challenging and significant for smart manufacturing.
| Category (12 classes) | 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 |
The PVELAD-2021 Datasets Request Form is available here.
All researchers need to follow the instructions below to access the datasets.
-
Download and fill the Industrial Datasets Request Form (MUST be hand signed with date). Please use institutional email address(es). Commercial emails such as Gmail and QQmail are NOT allowed.
-
Email the signed Industrial Datasets Request Form to Subinyi@buaa.edu.cn
-
Note that If you want to download through google disk, please send me your google email.
-
The dataset is jointly released by Hebei University of Technology and Beihang University.
[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.