论文阅读:《Bootstrap Your Own LatentA New Approach to Self-Supervised Learning》论文地址:https://...
论文阅读:《Bootstrap Your Own LatentA New Approach to Self-Supervised Learning》论文地址:https://...
接着前面的文章说到的transformer,本篇将要介绍在图像中如何将transformer运用到图片分类中去的。我们知道CNN具有平移不变形,但是transformer基于...
论文阅读:《AN IMAGE IS WORTH 16X16 WORDS:TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE》论文地址:ht...
论文阅读:《Improved Baselines with Momentum Contrastive Learning》论文地址:https://arxiv.org/abs/...
论文阅读:《A Simple Framework for Contrastive Learning of Visual Representations》论文地址:https:...
论文阅读:《Momentum Contrast for Unsupervised Visual Representation Learning》论文地址:https://ar...
论文阅读:《Batch Normalization: Accelerating Deep Network Training by Reducing Internal Cova...
图像的几何变换主要包括:平移、缩放、旋转、仿射、透视等等。图像变换是建立在矩阵运算基础上的,通过矩阵运算可以很快的找到不同图像的对应关系。理解变换的原理需要理解变换的构造方法...
概率论 统计学 常见数字特征 数学期望(均值) 在概率论和统计学中,数学期望(mean)(或均值,亦简称期望)是试验中每次可能结果的概率乘以其结果的总和,是最基本的数学特征之...
一会儿机器学习,一会儿深度学习,一会儿强化学习,一会儿深度强化学习,晕头转向All of a sudden every one is talking about them –...
数据集偏移&领域偏移 Dataset Shift&Domain Shift 将数据集的gap(differences between datasets)这个问题抽象出来...
论文阅读:O-GAN: Extremely Concise Approach for Auto-Encoding Generative Adversarial Network...