Attention residual network with multi-scale convolution branch for efficient solar photovoltaic module defect classification

Oluwatoyosi F. Bamisile

doi:10.1016/j.solener.2026.114323

Attention residual network with multi-scale convolution branch for efficient solar photovoltaic module defect classification

2026-01-13

0

PRE

AI

O

Oluwatoyosi F. Bamisile

S

She Kun *

C

Chiagoziem C. Ukwuoma

D

Dara Thomas

C

Chukwuebuka Joseph Ejiyi

DOI：10.1016/j.solener.2026.114323

Original

Original request for help

Save

Abstract

En 中文

• Multi-scale conv branches capture defects of varying sizes effectively. • Sequential Fusion Attention refines channel and spatial features for robust detection. • Residual learning stabilises deep training and preserves textural details. • Parallel conv pathways with diverse kernels enhance feature extraction. • Adaptive feature concatenation improves hierarchical defect detection.

Attention residual network with multi-scale convolution branch for efficient solar photovoltaic module defect classification

Abstract

Journal

Organization