End-to-End Fusion Network of Deep and Hand-Crafted Features for Small Object Detection

Recent advances in deep learning have enabled state-of-the-art performance detecting medium and large-size objects. However, small object detection remains challenging primarily due to the scarcity of information. This paper proposes an end-to-end fusion network that integrates hand-crafted features address this limitation. A module based on semantic context information is designed enhance feature discrimination ability. Additionally, we introduce a kind feature-contrast loss incorporate prior knowledge into according contrastive learning. Experiments MS COCO (34.4% ${\mathrm {A}}{{\mathrm {P}}_{\mathrm {S}}}$ ) PASCAL VOC (85.9% mAP) datasets demonstrate our approach achieves improved accuracy over previous methods, especially for