Pedestrian attribute recognition method based on image and attribute joint guidance

Abstract

A pedestrian attribute recognition method based on image and attribute joint guidance includes: obtaining image features and attribute features; build an image-attribute mutual guidance mechanism, including constructing the features of image guiding attributes and pedestrian attributes guiding images; inputting the feature of image guide attribute and the feature of pedestrian attribute guide image into a short-and long-term memory model, and obtaining a pedestrian attribute recognition model in which the image and attribute guide each other; obtaining the pedestrian attribute result and optimizing the pedestrian attribute result by continuous training using cross-entropy objective function. The method can make the model learn the relationship between the image feature and the attribute feature better through the mutual guidance between the image feature and the attribute feature, and adds the attention mechanism in the attribute recognition process, so that the model can further improve the distribution of the two features to improve the accuracy of the pedestrian attribute recognition.

Description

technical field [0001] The invention relates to a pedestrian attribute recognition method. In particular, it involves a pedestrian attribute recognition method based on image and attribute joint guidance. Background technique [0002] In order to protect the safety of people's lives and property, most countries in the world have launched development plans for safe cities, installing millions of surveillance cameras in different corners of the city, and these cameras are acquiring image information all the time and then analyzing it Research. Pedestrians are important objects of concern in video surveillance. Attribute identification is mainly to analyze pedestrian attributes in real monitoring scenarios, such as gender, age, clothing type, etc., which plays an important role in pedestrian retrieval and pedestrian re-identification in the field of video surveillance. Improving the recognition effect of pedestrian attributes can effectively identify people or things that ar...

AI Technical Summary

Problems solved by technology

[0004] Although some progress has been made in the recognition of pedestrian attributes, due to the low resolution efficiency of pedestrian images (especially images taken at long distances), the changes in posture, angle, illumination, etc. are very large, and the relationship between pedestrian attributes is relatively complex, so the recognition effect is still low. It needs to be improved. The main reason is that in the process of attribute recognition, images and attributes are trained separately, and the relationship between the two is not explored. Some improved methods also improve recognition performance by improving visual optimization.

For example, WPAL-network and JRL-network both optimize the image features. These methods can indeed improve the recognition performance, but they ignore the effect of attribute features on the overall recognition effect.

Images