YOLOv11-LWMB: A Hybrid YOLOv11 Framework for the Detection of ' Bull's-Eye ' Effect in Seismic Data

Accurate identification of ring-shaped amplitude artifacts, commonly referred to as the “Bull’s-Eye” effect, remains a key challenge in seismic inversion due to low-frequency deficiencies and complex subsurface heterogeneity. To address this issue, a hybrid lightweight framework named YOLOv11-LWMB is proposed for the automatic detection of Bull’s-Eye anomalies in post-stack seismic images. The model introduces a multi-component design in which MobileNetV4 enhances feature extraction efficiency, C3k2-WTConv expands the receptive field and strengthens low-frequency texture perception, LSKAttention adaptively models spatial context through large-kernel decomposition, and BiFPN realizes bidirectional multi-scale fusion for consistent anomaly localization. Extensive experiments on real seismic datasets demonstrate that YOLOv11-LWMB achieves substantial improvements over the original YOLOv11 baseline, with increases of 5.9 % in precision, 10.0 % in recall, and 12.8 % in mAP@0.5, while maintaining fast inference and low computational demand. These results confirm the model’s robustness in detecting weak and blurred seismic anomalies and highlight its potential for intelligent quality control and automatic interpretation in seismic inversion workflows.