Leukocyte detection method based on multi-scale fusion and deformable self-attention

The invention discloses a leukocyte detection method based on multi-scale fusion and deformable self-attention. The leukocyte detection method comprises the following steps: S1, acquiring a leukocyte target detection data set and preprocessing the leukocyte target detection data set; s2, an MFDS-DETR model is constructed and trained; S3, features are extracted through a backbone network to obtain a feature map; s4, obtaining a multi-scale fusion feature vector through a high-level screening feature fusion pyramid; s5, obtaining a final output vector through a deformable self-attention module; s6, taking the final output vector output in the step S5 as input, and extracting global features through an encoder; and S7, taking the extracted global features as input, and obtaining a final target detection result through a decoder. According to the method, the high-level semantic features serve as weights to screen the low-level features, the screened features and the high-level semantic features are added point by