专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明专利

AU2021102450A4 METHODOLOGY TO DETECT CRIME USING COMPUTER VISION AND DEEP LEARNING FOR SAFER NATION 未知
公开(公告)号：AU2021102450A4
公开(公告)日：2021-06-24
申请号：AU2021102450
申请日：2021-05-10
申请人： CHOUDHARI VINOD A , DIXIT ASHUTOSH , GHORPADE NAVEEN , MOHAN M S KIRAN , RANGA SHIVAPRAKASH , S VANITHA , SHUBHAM KUMAR , SINGH SHANTANU , SUDHIR BALE AJAY , TIWARI SUBHASHISH , TRIPATHI ABHISHEK
发明人： GHORPADE NAVEEN , SUDHIR BALE AJAY , TIWARI SUBHASHISH , RANGA SHIVAPRAKASH , SINGH SHANTANU , S VANITHA , TRIPATHI ABHISHEK , DIXIT ASHUTOSH , CHOUDHARI VINOD A , SHUBHAM KUMAR , MOHAN M S KIRAN
IPC分类号： G08B13/196 , G06K9/00 , G06N3/08 , G06Q50/26 , H04N7/18
摘要： METHODOLOGY TO DETECT CRIME USING COMPUTER VISION AND DEEP LEARNING FOR SAFER NATION Abstract: We live in unsafe environments. One of the utmost needs in today's society is one of security for both the humans and the other man-made resources. In this invention, we intended to mitigate the risk to life and that of human resources to a minimum by making use of the technologies that are already available to use. We Intend to employ the vast communication networks, i.e., the internet, the already in place surveillance devices such as CCTV cameras and computer Vision to Raise alerts to the nearest of the Emergency units like ambulance, fire station, police station, etc. for fast and efficient dispensing aid. In semantic pattern recognition, the role of images have been radically boosted by deep learning models with their explosive performance and growth of memory space in computing machines, besides the recent specialization. Personality traits are manifested by facial images and the author's individual characteristics are revealed by some textual post on social media. And so, there have been a few attempts such that personality traits are inferred from facial images. Now we are going to explore another level of image understanding keeping this ultimate goal in mind by using deep learning and computer vision to infer criminal tendencies from facial images. So there are two learning models which can be applied to discriminate between criminal and non-criminal facial images, a convolutional neural network (CNN) and a standard feedforward neural network (SNN). We can thus report for both models the training and test accuracies and the confusion matrix as well on a set of facial images by using tenfold cross-validation. In learning to reach the best test accuracy, it was found that the Convolutional neural network was more consistent than the SNN, with an 8% higher than the SNN's test. There is very little to gender bias in the classifier. This was found as no meaningful discrepancies in learning consistencies or classification accuracies were observed on trying with male gender facial images. Finally, in order to classify the two sets of images, the CNN takes advantage the shape of the face, eyebrows, top of the eye, pupils, nostrils, and lips by dissecting and visualizing convolutional layers. Database Emergency Server Processing Unit Af obileEmergency Emergency Unit Unit Fig 1: Work flow of the Model Fig 2: Design of the Model

IPRDB

热门服务

关于我们

友情链接

联系方式