You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

This is text-to-image generation model based on “stabilityai/stable-diffusion-2-base” for generating electroluminescence (EL) images of normal and defective PV modules. It can generate EL images data with multiple defects and diverse representations. This version is trained for 9000 steps. It is first trained up to 6000 steps and that version is available as "mwaqarakram/PVEL-Text-to-image-Generator-1" at the link: https://huggingface.co/mwaqarakram/PVEL-Text-to-image-Generator-1

Research study: This model is the result of a research study “M. W. Akram, J. Bai. 2026. Synergistic Integration of Text-to-Image Generation and Deep Learning for Photovoltaic System Inspection. Energy, 342 (139611). https://doi.org/10.1016/j.energy.2025.139611 ”. The model is trained with electroluminescence images of PV modules having resolution 512x512.

Training Data: The data used to train this model was captured from photovoltaic (PV) modules of different types and makes (brands) exposed to accelerated ageing in climate chambers and other field loading simulation setups like thermal cycling exposure chamber, potential induced degradation chamber, accelerated ultraviolet exposure chamber, static and dynamic mechanical loading simulation setups, and hail simulation setup to project real field operations. There were total 623 electroluminescence (EL) images in that data consisting of 670, 2160, 176, 7624, 1654, 1980 and 1422 instances of black cell, black edge, break, contamination, crack, finger interruption and low cell defects respectively. This data was then labeled with textual descriptions (text-image pairs). The labels include descriptions based on number of defects (quantities or items), type, severity, and/or other variations.

Other Details: Developed by: Muhammad Waqar Akram Model type: Diffusion-based text-to-image generation model Language: English

Cite as: The above research paper

Generated dataset: The electroluminescence images of PV modules generated from this model (at different stages of training) are available on Kaggle with name “Text-to-image generated PV module EL images” at following link: https://www.kaggle.com/datasets/waqarakram/text-to-image-generated-pv-module-el-images

The model is intended for research purposes only.

Downloads last month: 1