Pedestrian attribute recognition system and method based on multi-layer feature learning

Abstract

The invention discloses a pedestrian attribute recognition system and method based on multi-layer feature learning. The system includes a bottom-up feature extraction module, a top-down feature fusion module, a feature prediction module, a multi-layer prediction fusion module and a test module. , the specific steps of this method are: process the image layer by layer from bottom to top to obtain multi-layer features; fuse the features of adjacent layers from top to bottom, compress the channel of the feature map obtained by the higher layer, and combine it with the upper layer The sampled feature map is subjected to feature fusion and channel dimensionality reduction, and the current layer features are output; the fused features and the extracted uppermost layer features are passed through the maximum pooling layer and the fully connected layer to obtain preliminary prediction results at different levels; The preliminary prediction results are superimposed, and each attribute of each layer of prediction is given a corresponding weight value to obtain the final prediction result; the prediction result corresponding to the picture is extracted, and the results of each index are calculated. The present invention learns a set of specific weights for each attribute based on the predicted value obtained by the fused features, so that each attribute can better use multi-layer features to obtain better recognition effect.

Description

technical field [0001] The invention relates to the field of computer vision based on deep learning, in particular to a pedestrian attribute recognition system and method based on multi-layer feature learning. Background technique [0002] Pedestrian attribute recognition, also known as human body attribute recognition, is to identify the attributes of pedestrians in pictures, such as hair color, hair length, clothing type, clothing color, etc., which is a multi-label classification problem. Pedestrian attribute recognition is widely used in pedestrian re-identification and pedestrian detection under the monitoring system, and pedestrian attributes can be used to assist these tasks. In the field of pedestrian attribute recognition, most of the existing methods are based on the implementation of deep learning. For example, the attention mechanism is used to generate the attention map and applied to each layer of the neural network to obtain different features, that is, the at...

AI Technical Summary

Problems solved by technology

[0003] Existing technologies are not very good at learning the required features for each attribute, because the multi-layer feature fusion done in these works is only done at the feature level, and most of the operations are splicing feature vectors. It cannot make better use of and integrate multi-layer features, and there is no corresponding learning for each specific attribute to pay attention to the features of each layer

Images