Dense sampling and detail enhancement network: Improved small object detection based on dense sampling and detail enhancement

Small objects only occupy a few pixels in an image, which results low performance of small object detection for existing algorithms. Therefore, the authors propose dense sampling and detail enhancement network (DSDE-Net) to address this issue. The contains module used increase resolution feature maps expand receptive field, includes atrous spatial pyramid pooling coordinate attention mechanism systematically process maps. Simultaneously, introduce branch that edge detailed information generate through Gaussian filtering compensate loss occurs extraction process. experimental demonstrate proposed outperformed related methods. Compared with state-of-the-art algorithm DetectoRS, it effectively achieves approximately 4.6% improvement on minicoco2021 dataset 4.2% remotely sensed VisDrone.