A target detection method, device, electronic equipment and storage medium

Abstract

The present application discloses a target detection method, which includes: determining a random pseudo-target according to the obtained unlabeled data set, and determining the labeled data of the random pseudo-target; using the unlabeled data set and labeled data to train to obtain a pseudo-target detection model; using The manually labeled data set continues to train the pseudo-target detection model to obtain the target detection model; obtain the data of the target to be detected, and input the data of the target to be detected into the target detection model to obtain the detection result of the target to be detected. This method makes full use of the unlabeled data set, and obtains a pseudo-target detection model based on the unlabeled data set, then inputs the artificially labeled data set into the pseudo-target detection model to obtain a target detection model, and finally obtains the detection result, which can improve target detection. model effect. The present application also provides a target detection device, an electronic device and a storage medium, which have the above beneficial effects.

Description

technical field [0001] The present application relates to the technical field of target detection, in particular to a target detection method, device, electronic equipment and storage medium. Background technique [0002] At present, computer vision technology and related deep learning are widely used, for example, in the microscopic field of view under a microscope related to medical diagnosis. Among them, supervised learning means that the data used for training contains information manually labeled, and self-supervised learning means extracting relevant facts from the data itself as labels without manual labeling. [0003] In related technologies, most of the target detection applications are supervised training based on labeled data, that is, the position, size and target category of the target to be detected are marked in the image, and more traditional flipping and deformation are adopted during training based on the fact that the labeled data is not detected. The sev...

AI Technical Summary

Problems solved by technology

[0003] In related technologies, most of the target detection applications are supervised training based on labeled data, that is, the position, size and target category of the target to be detected are marked in the image, and more traditional flipping and deformation are adopted during training based on the fact that the labeled data is not detected. The severely damaged method performs data enhancement, but there is a problem of low data utilization, that is, only artificially labeled data is used for training, and the performance of the target detection model is relatively low, and a large amount of manually labeled data is often required in specific applications. Labeling is also expensive

Images