Datasets:
PV_Multi_Defect-corrected: 1044 images / 3403 boxes / 5 classes. Removed 61 byte-identical duplicates and 3 orphan/empty files, kept the near-duplicate cluster on one side of a seeded 70/15/15 stratified split, cleaned VOC metadata (personal hard-coded path + bogus folder), added YOLO TXT + classes.txt + data.yaml + CSVs
4d5eac6 | PV_Multi_Defect-corrected — LICENSE | |
| ==================================== | |
| This file covers TWO different things. Please read both parts. | |
| -------------------------------------------------------------------------------- | |
| PART 1 — The modifications made in this release | |
| -------------------------------------------------------------------------------- | |
| The changes made to produce "PV_Multi_Defect-corrected" are released under the | |
| Creative Commons Attribution 4.0 International License (CC BY 4.0) | |
| Full legal text: https://creativecommons.org/licenses/by/4.0/legalcode | |
| These modifications are: | |
| * removal of 61 byte-identical duplicate images (61 redundant copies deleted; | |
| one copy of each pair was kept), | |
| * removal of 2 images that had no annotation file (img3059, img9513) and of 1 | |
| annotation file that contained no objects (img7863), | |
| * grouping of the single near-duplicate cluster (12 perceptually identical | |
| images) so it is assigned entirely to one side of the split and cannot leak | |
| across train/val/test, | |
| * addition of a seeded, class-stratified train/val/test split (723 / 155 / 166) | |
| — upstream ships no split at all, | |
| * clean-up of the Pascal VOC metadata: the upstream XML carried a bogus | |
| <folder> value ("new" / "result2"), a hard-coded local <path> exposing a | |
| personal Windows path (C:\Users\86151\Desktop\...), and <database>Unknown</database>, | |
| * conversion of the annotations to YOLO format alongside Pascal VOC, | |
| * regeneration of classes.txt, data.yaml, the per-split CSV tables and this | |
| documentation. | |
| NO bounding box coordinate was altered. Only the packaging, the split and the | |
| file metadata changed. The image pixels are byte-for-byte identical to upstream. | |
| You are free to share and adapt these modifications for any purpose, including | |
| commercially, provided you give appropriate credit. | |
| -------------------------------------------------------------------------------- | |
| PART 2 — The underlying PV-Multi-Defect data (IMPORTANT) | |
| -------------------------------------------------------------------------------- | |
| The underlying images and annotations originate from the PV-Multi-Defect dataset: | |
| PV-Multi-Defect Dataset | |
| Photovoltaic (PV) panel multi-defect detection, 5 defect classes. | |
| GitHub: https://github.com/houhou34/PV-Multi-Defect | |
| Reference: Li, L., Wang, Z. & Zhang, T. (2023). "GBH-YOLOv5: Ghost | |
| Convolution with BottleneckCSP and Tiny Target Prediction Head Incorporating | |
| YOLOv5 for PV Panel Defect Detection." Electronics, 12(3), 561. | |
| https://doi.org/10.3390/electronics12030561 | |
| NO LICENCE IS STATED FOR THE DATASET. | |
| * The GitHub repository ships no LICENSE file and no terms of use. | |
| * The associated article is published open access by MDPI. | |
| Consequently, the copyright status of the underlying PV-Multi-Defect images and | |
| annotations is UNDETERMINED. No licence is granted for them by this file. | |
| >>> Before any commercial use or redistribution of the image data, you must | |
| >>> determine the applicable terms from the original authors | |
| >>> (contact: zfwang@ccnu.edu.cn). | |
| -------------------------------------------------------------------------------- | |
| If you are the rights holder of the PV-Multi-Defect dataset and would like this | |
| release changed or removed, please open an issue and it will be actioned promptly. | |
| -------------------------------------------------------------------------------- | |