Object part detection method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an object part detection method and device, electronic equipment and a storage medium, and the method comprises the steps: building a corresponding prior loss function based on a logic association relationship between target part categories in a detection task, and carrying out the detection of target part types based on the built prior loss function in a model training process, and enhancing the association relationship between the target part categories, so that the target detection model can determine the corresponding association constraint information between the to-be-output detection part categories according to the above steps, thereby optimizing the false detection object part and the missing detection object part in the detection result, and improving the accuracy of the detection result. The accuracy of the detection result output by the target detection model is improved, further, in the model training process, additional data dimensions are not added, multi-round iterative training is performed on the model based on construction of a corresponding prior loss function, the deployment cost is lower, the generalization ability of the target detection model is ensured, and the target detection efficiency is improved. And the deployability of the target detection model is improved.

Description

technical field [0001] The present invention relates to the field of computer vision, and in particular, to a method, device, electronic device and storage medium for detecting object parts. Background technique [0002] Object detection technology (Object Detection) is a computer vision processing technology based on deep learning, neural network and other technical means to determine and obtain the respective categories and relative positions of each target object (object) in the specified target image. , target tracking and other fields play a vital role. [0003] For example, in the process of recognizing a target image, the target detection model specifically performs the following four types of tasks: Classification, Location, Detection, and Segmentation. Specifically, in the process of model training, it is usually necessary to generate corresponding detection frames containing the specified candidate objects according to each candidate feature extracted from the sam...

AI Technical Summary

Problems solved by technology

[0004] 1. Low detection accuracy

[0005] In related technologies, in order to ensure that the training of the target detection model is sufficient, it is usually necessary to perform data enhancement on the sample images during the training process, so as to respectively simulate the sample images corresponding to different light intensities, different resolutions and other complex situations. However, in some special In some cases, for example, when the light intensity of the sample image is strong and the resolution is low, since the candidate features extracted by the neural network from the sample image are not accurate enough, when based on the preset target feature library, for the extracted candidate features When performing classification judgments, the category of the corresponding candidate object is often mistaken for the category of other objects with high feature similarity, that is, the misdetection of the target category occurs, which affects the accuracy of the output results of the target detection model

[0006] 2. Weak model generalization ability

[0008] However, on the one hand, this method requires a lot of time and calculation costs. On the other hand, this training method easily makes it difficult for the target detection model to ensure the accuracy of target detection in target detection scenarios other than various training scenarios. , that is, the generalization ability of the target detection model is limited

Images