1. 几种常见噪声高斯噪声概率密度函数服从高斯分布的噪声。产生原因：1）图像传感器在拍摄时市场不够明亮、亮度不够均匀；2）电路各元器件自身噪声和相互影响；3）图像传感器长期工作，温度过高代码实现：def gasuss_noise(image,mean=0,var=0.001):'''手动添加高斯噪声mean : 均值var : 方差'''image = np.array(image/255, dt

OCR综述概览1. 文字识别指标为f1-scoreConf.NetTitleSVTIIIT5KICDAR13TPAMI2015CRNNAn end-to-end trainable neural network for image-based sequencerecognition and its application to scene text recognition0.8080.7820.8

目标检测各种Tricks总结

平滑 也称 模糊, 是一项简单且使用频率很高的图像处理方法。平滑处理的用途有很多， 但是在本教程中我们仅仅关注它减少噪声的功用。从滤波角度来讲，一般主要的目的都是为了实现对图像噪声的消除，增强图像的效果。均值滤波代码：def blur_demo(image):# 均值模糊dst = cv.blur(image,(3,3))# ksize是卷积核大小3行3列,列数越大模糊越大cv.imshow("b

OCR综述概览1. 文字识别指标为f1-scoreConf.NetTitleSVTIIIT5KICDAR13TPAMI2015CRNNAn end-to-end trainable neural network for image-based sequencerecognition and its application to scene text recognition0.8080.7820.8

OCR综述概览1. 文字识别指标为f1-scoreConf.NetTitleSVTIIIT5KICDAR13TPAMI2015CRNNAn end-to-end trainable neural network for image-based sequencerecognition and its application to scene text recognition0.8080.7820.8

人体姿态估计综述 2D 3D

mask转coco的json