Blood cell target detection based on improved YOLOv10

The detection of blood cell image has always been a very key branch in the field of target detection, which belongs to small target detection in a broad sense. Due to the high similarity between different cells in cell detection, the overlapping occlusion of different cells, and the lack of obvious characteristics of target cells in the detection of blood, the detection of target cells poses a greater challenge. In order to better solve the above problems, we propose a linear attention feature fusion network based on YOLOv10(LAFF-YOLOv10). In this paper, the state space attention module (SSAM) introduced to improve the detection quality of small target feature extraction. Secondly, a multilevel feature fusion module (MFFM) is proposed. Finally, the effectiveness, superiority, lightness and universality of the proposed method on private ZBCD, HBCD, and public BCCD data sets are verified. The experimental results showed that AFF-YOLOv10 was superior to baseline model (yolov10), and successfully improved the detection performance of blood cell images.