Aircraft detection based on improved DETR

With the extensive application of deep learning in the field of computer vision, Transformer-based object detection methods, particularly DETR (Detection Transformer), have become a research hotspot. However, DETR still faces challenges of low detection accuracy and efficiency when dealing with small objects and complex backgrounds. To enhance DETR's performance in aircraft detection, this paper proposes an improved DETR-based aircraft detection method, incorporating two innovative modifications: First, an ECA (Efficient Channel Attention) mechanism is introduced after CNN feature extraction to enhance the model's adaptive learning ability for important features. Second, the original encoder-decoder structure in DETR is replaced with a Swin Transformer, leveraging its local windowed self-attention mechanism to improve computational efficiency and accuracy. Experimental results demonstrate that the proposed method achieves higher detection accuracy and speed compared to traditional DETR in aircraft detection tasks, particularly showing better robustness and generalization capabilities under complex backgrounds. This research provides an effective improvement approach for Transformer-based object detection methods and offers new technical support for high-precision object detection tasks in practical applications.