PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards

HZDR - Hif_Exploration

Overview

Our primary focus is to enhance the non-invasive optical analysis of E-waste materials, specifically plastics and printed circuit boards (PCBs). We aim to develop a smart multisensor network that utilizes RGB cameras and hyperspectral imaging, along with other types of sensors, to improve the efficiency of the E-waste recycling industry. This involves providing both quantitative and qualitative information that aids in decision-making for subsequent sorting and processing.

Research Paper

This GitHub repository corresponds to the research paper titled "PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards." The paper introduces the first RGB-Hyperspectral Imaging (HSI) benchmark segmentation dataset for PCBs. You can access the paper here.

images/training_hsi.png

Dataset Details

The dataset includes:

• RGB images of 53 PCBs scanned with a high-resolution RGB camera (Teledyne Dalsa C4020).
• 53 hyperspectral data cubes of those PCBs scanned with Specim FX10 in the VNIR range.
• Two segmentation ground truth files: 'General' and 'Monoseg' for 4 classes of interest - 'others,' 'IC,' 'Capacitor,' and 'Connectors.'

Code Repository

The repository includes code to read, manipulate, and process large-scale data. It also provides examples of preprocessing and processing steps, such as dimensionality reduction (e.g., PCA) and image segmentation using 5 deep learning models.

Requirements

To use the codes without errors, install the libraries listed in the Requirements.txt file. The codes require at least 1 GPU to run and handle the data.

Usage

For detailed code instructions, please refer to the code documentation. More information about the methodology and experiments can be found in the paper here.

Data Access

To utilize the dataset, download it from the Rodare website: Rodare, or from Zenodo: Zenodo.

Contributions

All comments and contributions are welcomed. The repository can be forked, edited, and pushed to different branches for enhancements. Feel free to contact me directly at e.arbash@hzdr.de or via our website.

License

The code is licensed under the Apache-2.0 license. Any further development and application using this work should be opened and shared with the community.

Acknowledgment

The authors express their gratitude to EIT RawMaterials for funding the project ’RAMSES-4-CE’ (KIC RM 19262). Appreciation is extended to the European Regional Development Fund (EFRE) and the Land of Saxony for their support in funding the computational equipment under the project ’CirculAIre.’

Citation

When using the materials of the work and dataset, please cite it as follows: Word:

Arbash, Elias, Fuchs, Margret, Rasti, Behnood, Lorenz, Sandra, Ghamisi, Pedram, & Gloaguen, Richard. (2024). PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards (Version 1) [Data set]. Rodare. http://doi.org/10.14278/rodare.270

Latex:

@article{arbash2024pcb,
  title={PCB-Vision: A Multiscene RGB-Hyperspectral Benchmark Dataset of Printed Circuit Boards},
  author={Arbash, Elias and Fuchs, Margret and Rasti, Behnood and Lorenz, Sandra and Ghamisi, Pedram and Gloaguen, Richard},
  journal={arXiv preprint arXiv:2401.06528},
  year={2024}
}