What Are the Parking Lot System Technologies Behind the Urban Brain- TigerWong

The parking lot system uses content extraction network and style extraction network to extract features from content pictures and style pictures respectively, inspired by conditional confrontation generation network and style transfer learning. After fusing the two, images with corresponding content and style are obtained through the image generation network. Internet based infrastructure, rural brain. Make use of rich rural data resources to make an overall and real-time analysis of the countryside. Effectively allocate public resources with urban data resources, improve social governance from time to time, and promote rural sustainable development. Assembled into a mini e-book, today I will write three practical papers on "rural brain". The three excellent papers shared with you were included in the acmmm2017 summit papers, an important part of the technology in front of the rural brain.

Every technological revolution, Wang Jian, chairman of the technical committee of Alibaba group, once said. Will push forward rural civilization. In the age of steam engine, the sign of countryside was to build roads; In the era of electric power, the development of rural areas is to lay the power grid. In the Internet age, data has become an important resource. Villages need to build a data brain to improve civilization again. Find many problems to be solved through the "rural brain" project.

Parking lot system, such as how to accurately identify the flow of people and vehicle tracks, how to extract the object features in the three-dimensional space, etc. Through practice from time to time, find some better solutions, and put these methods into the actual scene to positively enhance the landing of the application. Video anomaly detection refers to detecting segments with abnormal state and behavior in a video. This paper aims to discuss how to solve the problem of "video anomaly" detection. Anomaly detection in real-world video scenes is a difficult problem, because "anomaly" itself is difficult to define, and there are chaotic backgrounds, objects and motions in the scene. How to use anomaly detection algorithm to help the system automatically find traffic accidents and suspicious pedestrians in traffic and security scenarios? In real life, it has a wide application prospect and high research value.

Inspired by the latest research results in the field of motion recognition, in this paper, I provide a method for the rural brain to monitor traffic abnormalities. A spatiotemporal self coding is designed for video anomaly detection, and a prediction error calculation method with decreasing weight is proposed. Through the evaluation of real traffic scenes, the algorithm has exceeded the previous best method in important indicators. Accurately find the same person in the pictures corresponding to other cameras. Pedestrian re recognition technology has very important scientific research and practical application value.Pedestrian re recognition refers to the picture of a pedestrian under a given camera. Recently, it has been widely used in transportation, security and other fields, which is of great significance to create a safe and intelligent village. At the same time, the similarity is extended to other levels. In this paper, I provide technical support for the identification and judgment of pedestrian flow trajectory. Combining the advantages of siames network and classification network model of deep learning.At present, the optimal result of retrieval accuracy has reached the highest level in the industry. One of the most basic models of smart village, license plate recognition model. We need to identify the license plates of all villages in the country. For example, I collect data mainly based on "Zhejiang a license plate is the main license plate. At this time, algorithms need to be used to automatically generate data.

The rise of in-depth learning in recent two years has greatly improved the performance of AI model, but it also brings huge data demand and labeling cost. In real scenes, it is often difficult to obtain more comprehensive, large, uniform and accurately labeled data, so let AI learn to generate data as a master Flow direction is one of the classical Gan models, but ordinary Gan models often have some problems, such as uncontrollable data generation, poor generalization and so on.