zae-bayern/elpv-dataset

What it solves

This project provides a benchmark dataset for the visual identification of defective solar cells. It addresses the need for standardized data to train and evaluate models that can automatically detect intrinsic or extrinsic defects that reduce the power efficiency of photovoltaic modules.

How it works

The repository provides a Python package that allows users to easily load a curated collection of 2,624 grayscale images (300x300 pixels) of solar cells. These images were extracted from high-resolution electroluminescence imagery of 44 different solar modules. The images have been normalized for size and perspective, and camera lens distortion has been removed.

Who it’s for

Researchers and developers working on computer vision and machine learning models for solar energy quality control and defect detection.

Highlights

• 2,624 annotated samples of functional and defective solar cells.
• Images are normalized for size and perspective with lens distortion removed.
• Each image is annotated with a defect probability (0 to 1) and the solar module type (mono- or polycrystalline).
• Easy installation via pip and simple API for loading the dataset.