Bridging the Data Gap: Spatially Conditioned Diffusion Model for Anomaly Generation in Photovoltaic Electroluminescence Images

TL;DR

This paper presents PV-DDPM, a novel spatially conditioned diffusion model for generating diverse PV electroluminescence images with anomalies, enhancing dataset diversity and improving anomaly detection performance.

Contribution

Introduces PV-DDPM, the first model to jointly generate multiple PV cell types with controlled anomalies, and creates an enriched dataset E-SCDD for better PV defect detection.

Findings

Generated images achieve low FID and KID scores.

Training AA-CLIP on E-SCDD improves detection metrics.

PV-DDPM enables controlled synthesis of diverse PV anomalies.

Abstract

Reliable anomaly detection in photovoltaic (PV) modules is critical for maintaining solar energy efficiency. However, developing robust computer vision models for PV inspection is constrained by the scarcity of large-scale, diverse, and balanced datasets. This study introduces PV-DDPM, a spatially conditioned denoising diffusion probabilistic model that generates anomalous electroluminescence (EL) images across four PV cell types: multi-crystalline silicon (multi-c-Si), mono-crystalline silicon (mono-c-Si), half-cut multi-c-Si, and interdigitated back contact (IBC) with dogbone interconnect. PV-DDPM enables controlled synthesis of single-defect and multi-defect scenarios by conditioning on binary masks representing structural features and defect positions. To the best of our knowledge, this is the first framework that jointly models multiple PV cell types while supporting simultaneous generation of diverse anomaly types. We also introduce E-SCDD, an enhanced version of the SCDD dataset, comprising 1,000 pixel-wise annotated EL images spanning 30 semantic classes, and 1,768 unlabeled synthetic samples. Quantitative evaluation shows our generated images achieve a Fr\'echet Inception Distance (FID) of 4.10 and Kernel Inception Distance (KID) of 0.0023 $\pm$ 0.0007 across all categories. Training the vision--language anomaly detection model AA-CLIP on E-SCDD, compared to the SCDD dataset, improves pixel-level AUC and average precision by 1.70 and 8.34 points, respectively.