Target detection deep learning model training method and target detection method

Abstract

The invention discloses a method for training a deep learning model in target detection based on deep learning, which comprises the following steps of: establishing a model training picture set whichcomprises a negative sample training set consisting of actual scene pictures which do not contain a detection target and a positive sample training set consisting of non-actual scene pictures which contain the detection target; and training a deep learning model by using the model training picture set and the information related to the detection target and the pictures in the model training picture set. According to the method, the loss of the background picture is considered when the loss is calculated, so that the false detection rate of the model on the background in an actual detection scene is greatly reduced. Strong robustness is provided for illumination change and scene change. The operation speed is high, real-time operation detection can be carried out on the CPU, and stability and high efficiency are achieved; hardware requirements are simple, and large-scale popularization is easy.

Description

technical field [0001] The present invention relates to a target detection method based on deep learning in computer vision image processing technology, in particular to a method for training a deep learning model in target detection based on deep learning, and a corresponding target detection method. Background technique [0002] As a classic topic in the field of computer vision image processing, target detection has important applications in autonomous driving, traffic monitoring, image retrieval, etc. Its purpose is to detect and classify specific objects that people are interested in from images or videos. Traditional object detection methods such as HOG, SIFT, etc. usually separate image feature extraction and classification processes. These methods first use the feature model to extract the relevant visual features of the image, and then use a classifier, such as SVM, for recognition. [0003] Since Professor Hinton proposed the theory of Deep Learning, more and more...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to address the deficiencies and defects that occur when the existing deep learning-based target detection technology is applied to actual scenes, especially for the complex actual application detection scene and the lack of actual scene training picture sets containing detection targets. Improve

Images