Reverse attention model based on multi-scale depth supervision

Abstract

The invention discloses a reverse attention model based on multi-scale depth supervision. The model comprises an input end, a multi-scale feature learning module, an attention mechanism module, a reverse attention mechanism module, a depth supervision module, a plurality of loss functions, a plurality of average pooling layers, a plurality of linear layers and branches. The multi-scale feature learning module is used for performing multi-scale learning and training on the depth features; the attention mechanism module is used for enhancing attention on local important feature information; the reverse attention mechanism module is used for changing the features inhibited by the attention mechanism module into emphasized features and complementing the attention mechanism; and the deep supervision module is used for correcting the attention accuracy of the attention mechanism module on the important features. A reverse attention mechanism module is provided, the problem of feature information loss caused by an attention mechanism is relieved, part of modules can be discarded in the test stage of the model, and the test efficiency is improved.

Description

technical field [0001] The present invention relates to the field of pedestrian re-identification, in particular to a reverse attention model based on multi-scale deep supervision. Background technique [0002] Pedestrian Re-Identification (PReID) is the task of automatically judging whether the pedestrians captured by different traffic cameras or the same traffic camera at different times are the same pedestrian. Due to its important role in the application of intelligent video surveillance systems, person re-identification has received extensive attention in the field of computer vision in recent years. The resolution of pedestrians captured in real scenes is low, and traditional biometric information cannot be obtained accurately. At present, this task mainly relies on the appearance characteristics of pedestrians for identification. However, there are differences in illumination, posture, angle of view and background in pictures of pedestrians captured in different scen...

AI Technical Summary

Problems solved by technology

However, when these features are reweighted, some features are emphasized, and the attention of other features is weakened, resulting in the loss of some important feature information.

The network model based on multi-scale feature learning often embeds the multi-scale feature learning module into the feature extraction network. Although this embedding operation can improve the feature learning ability of the model to a certain extent, it will increase the complexity of the network model, so it is urgently needed. A model that can solve the problems in the prior art

Images