Tiny-YOLOv3-based pneumonia image target detection method

According to the technical scheme, the Tiny-YOLOv3-based pneumonia image target detection method comprises the following steps that (1) an original XML file is analyzed into TXT, and each row of the TXT file specifically comprises an image absolute path, coordinate values of the upper left corner and the lower right corner of a pneumonia focus and the probability that the focus exists; 2) the training data are trained by using a target detection model Tiny-YOLOv3, and finally a predicted value is obtained; (3) the difference between the position information and the probability information in the obtained predicted value and the position information and the probability information corresponding to the real value in the training set is processed through a loss function, and finally the convolution kernel parameter of the convolution operation in the feature extraction network is adjusted according to the processing result of the loss function; and 4) a to-be-detected pneumonia image is input to the trained target detection model, and the target detection model outputs a detection result which specifically comprises a coordinate value and a probability value. The method improves the model training precision, and enables the prediction result to be more accurate.