VanillaNet-Enhanced YOLO11 for Efficient and Accurate Aircraft Surface Damage Detection

This paper introduces VanillaNet-Enhanced YOLO11, a deep learning-based model designed for efficient and accurate aircraft surface damage detection. Our approach enhances the YOLO11 framework through a series of innovations, including depthwise separable convolutions, cross-stage partial connections, and advanced feature extraction modules from VanillaNet. Notable innovations include C3K2 (Cross-Stage Kinematic Kernel) for improved feature extraction, SPFF (Spatial Pyramid Fast Fusion) for better object size detection across varying scales, and C2PSA (Channel-Spatial Attention) for improved fusion of channel and spatial information. These modules work together to significantly boost detection performance, achieving superior accuracy and computational efficiency. Experimental results show that VanillaNet-Enhanced YOLO11 outperforms existing models in both speed and precision, offering a reliable solution for real-time aircraft inspection and contributing to enhanced aviation safety and operational efficiency.