Analysis of reversing image recognition problems based on lightweight model

Abstract

In recent years, people's living quality has improved, so the price of cars has fallen, and the number of cars in the world has been on the rise. In cities where land is expensive, there are fewer places to park, fewer parking Spaces, and parking itself is a difficult technology to learn. Therefore, the number of accidents caused by parking increases year by year. It is urgent to solve the safety problem of parking. Although some auxiliary astern tools have emerged as The Times require, these tools still have their shortcomings. For example, the commonly used astern radar and astern image cannot see objects behind, and astern image is a wide-angle lens that makes it difficult for the driver to judge the distance between the car and obstacles. In this environment, the current situation needs to be improved. In order to improve the accuracy of obstacle judgment, this paper chooses semantic segmentation as the function. This paper chooses lightweight model to speed up the process of deliver information, such as ShuffleNetV2.