Object of interest extraction based on deep learning

With the rapid development of the information age, more and more image messy information appear in our lives. However, people prefer to focus on the information they are interested in. Based on this, we propose a method for extracting objects of interest from the interference information using U-net network. To achieve this goal, we specially design the dataset that the labeled images only retain objects of interest, so that the network model only needs to learn the feature information of the object of interest related to the task, which can extract and preserve the feature information of the most relevant objects in the different scene. The objects of interest can be reconstructed in different scenarios under small self-built datasets. The method avoids processing the global information of all objects in the scene, greatly reducing the storage and transmission of useless information, and will have far-reaching application prospects in object recognition, object classification, etc.