Neural network for drawing multi-label identification and related method, media and device

Abstract

The invention discloses a neural network for drawing multi-label identification and a related method, a medium and a device. The neural network of the embodiment of the invention comprises a residualattention network which is used for receiving a picture image and learning an attention area of the image so as to output a feature map. The content label network is used for receiving the feature mapand outputting a prediction probability of the content label. The topic label network is used for receiving the feature map and outputting a prediction probability of topic labels. And the category label network is used for receiving the feature map and outputting the prediction probability of the category labels. According to the embodiment, the content multi-label identification, the subject multi-label identification and the category single-label identification of the painting image can be realized.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a neural network for multi-label recognition of paintings, a training method using the neural network, a method for multi-label recognition using the neural network, a storage medium and a computer device. Background technique [0002] Deep learning is one of the most important breakthroughs in artificial intelligence in the last decade. It has achieved great success in speech recognition, natural language processing, computer vision, image and video analysis, multimedia and many other fields. On the ImageNet dataset, the top-5error of ResNet is only 3.75%, which has greatly improved the index compared with traditional recognition methods. Convolutional neural network has powerful learning ability and efficient feature expression ability, and has achieved good results in single-label recognition. However, a real image contains more than one scene and object, so the pro...

AI Technical Summary

Problems solved by technology

[0003] The current existing methods are all based on ordinary photos to generate corresponding content tags or scene tags, without the characteristics of paintings (requires multiple types of tags, including multi-label and single-label; and ordinary photo image recognition does not require multiple similar paintings. class labels) to generate labels, and there is no method of putting single-label and multi-label generation in one network and generating labels at the same time