Cervical Cancer Cell Image Detection Method Based on Improved YOLOv4

Aiming at the problem of frequent false and missed detection due to the dense and scattered distribution of cancer cells in the pathological images of cervical cancer cells, the complex image background, and the rich proportion of small objects, an improved image detection method of cervical cancer cells is proposed by combining the advantage of convolutional neural network and the transformer on the basis of the YOLOv4 network. Firstly, this method introduces the CoT module into the backbone feature extraction network of YOLOv4 to improve the global information acquisition ability of the model and the visual expression ability of the model. Secondly, according to the characteristics of the cervical cancer cell object detection dataset, data augmentation strategies such as Mosaic are used to enrich the dataset and improve the generalization ability of the model. Lastly, The improved method proposed is used to perform experiments on cervical cancer cell image dataset, and the corresponding results are obtained. The results show that the improved network model is significantly improved, which significantly improves the detection ability of cervical cancer cells and fully verifies the effectiveness of the algorithm proposed in this paper.