An image abstract generation method based on object detection

Abstract

The invention discloses an image abstract generation method based on object detection, which comprises the following steps: feature extraction of image local area, attention mechanism system and imagedescription generation. The feature extraction of image local area is connected with the attention mechanism system, and the attention mechanism system is connected with the image description generation. The image local region feature extraction adopts Faster RCNN detection method to extract image region feature, and Faster RCNN consists of three parts: original image feature extraction, RPN network and ROI pooling. The present invention applies the object detection algorithm to the image summary task, and uses the local area image feature to represent the whole image, which greatly reduces the calculation amount and the waste of resources without losing the image information. The invention expands the research thought of the image summary task, and has certain reference value for the research of the image summary task.

Description

technical field [0001] The invention relates to the technical field of computer image processing, in particular to an image summary generation method based on target detection. Background technique [0002] Due to the rapid development of artificial intelligence, the use of deep learning methods to solve problems has become a hot topic. Image summarization is a comprehensive problem that combines computer vision and natural language processing. It is similar to translating an image into a description. For humans, this task is very easy, but it is very challenging for machines. The machine not only needs to use the model to understand the content of the picture, but also needs to use natural language to express the relationship between them. In addition, the model also needs It is not easy for machines that cannot think independently to be able to capture the semantic information of images and generate human-readable sentences. The image summarization task is of great signi...

AI Technical Summary

Problems solved by technology

[0006] The above-mentioned methods have their own advantages, but their shortcomings are also very obvious. The template-based method has a single sentence pattern, and the method based on the transfer title generation strategy is very inflexible. Although the traditional deep learning method is effective, However, the global information of the entire image is used, resulting in a waste of resources, which is a great consumption of computer hardware.

Images