arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2011.14603cs.CVcs.LG
收藏

REaL:基于欧氏空间和似然估计的实时的人脸检测与识别

REaL: Real-time Face Detection and Recognition Using Euclidean Space and Likelihood Estimation

Nitte Meenakshi Institute of Technology(尼特·米纳克希理工学院);Department of Information Science and Engineering(信息科学与工程学院)

Sandesh Ramesh, Manoj Kumar M, K Aditya Shastry

摘要:

准确地检测和识别人脸一直是一个挑战。区分面部特征、训练图像和快速生成结果需要大量的计算。我们在本文中提出的REaL系统讨论了它的功能和方法,可以在很短的时间内进行计算。在真实图像上进行了REaL实验,识别率是有希望的。该系统还成功地将非人类物体从其计算中移除。该系统使用本地数据库来存储捕获的图像,并频繁地向神经网络提供信息。捕获的图像被自动裁剪以去除不需要的噪声。该系统计算欧拉角和面部是否在微笑、左眼和右眼是否睁开的概率。

英文摘要:

Detecting and recognizing faces accurately has always been a challenge. Differentiating facial features, training images, and producing quick results require a lot of computation. The REaL system we have proposed in this paper discusses its functioning and ways in which computations can be carried out in a short period. REaL experiments are carried out on live images and the recognition rates are promising. The system is also successful in removing non-human objects from its calculations. The system uses a local database to store captured images and feeds the neural network frequently. The captured images are cropped automatically to remove unwanted noise. The system calculates the Euler angles and the probability of whether the face is smiling, has its left eye, and right eyes open or not.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑