A pedestrian reidentification data enhancement method based on stochastic linear interpolation

Abstract

The invention discloses a pedestrian re-identification data enhancement method based on stochastic linear interpolation, A stochastic linear interpolation method is used to increase the number of samples in the original pedestrian rerecognition dataset, and a large number of pedestrian images with different occlusion levels are used to obtain more abundant pedestrian data distribution informationin the dataset. Then, by using the benchmark depth learning network model to learn the enhanced data set, the generalization ability of the model is improved, and the Rank-1 matching erro of pedestrian recognition is reduced.

Description

technical field [0001] The invention relates to the fields of video monitoring and data processing, in particular to a pedestrian re-identification data enhancement method based on random linear interpolation. Background technique [0002] With the rapid development of deep learning, more and more convolutional neural network models can effectively handle computer vision tasks such as image classification and target detection. However, in order to ensure the generalization ability of large convolutional network models for small-scale data, especially However, when identifying pedestrians in different camera perspectives, we are facing the challenge of small data sets, and data enhancement methods are very important. The traditional data enhancement method is to enhance a single data, including algorithms such as random cropping, random flipping and random erasing, but on the pedestrian re-identification dataset, considering that the convolutional network model needs to have ...

AI Technical Summary

Problems solved by technology

[0004] When training on the pedestrian re-identification data set, the second method will work better, but the current more advanced method random erasing only considers the change of the pixel value of the sample itself, which will make the model less sensitive to the data distribution in the data set. There are limitations in the learning of pedestrian re-identification. Further improving the learning ability of the convolutional neural network model for the data distribution in the pedestrian re-identification dataset has become an urgent problem to be solved in the pedestrian re-identification.

Images