TE-YOLOF: Tiny and efficient YOLOF for blood cell detection

Blood cell detection in microscopic images is an essential branch of medical image processing research. The research and application computer vision algorithms this field are more concerned about the trade-off between accuracy model complexity. FED detector modified by YOLOv3 a representative light-weight to detect blood objects such as red cells, white cells platelets. But precision relatively low compared with platelets due imbalance distribution different types cells. In research, we propose based on YOLOF address problem, order achieve overall improvement precision. This object called TE-YOLOF, Tiny Efficient YOLOF. Model light-weighting accomplished excellent feature extraction capabilities EfficientNet backbone ability Depthwise Separable Convolution reduce number parameters while maintaining Furthermore, Mish activation function employed increase Extensive experiments BCCD dataset prove effectiveness proposed model, which can higher less than FED. TE-YOLOF also effective other cross-domain experiments.