![](https://img-blog.csdnimg.cn/20201014180756927.png?x-oss-process=image/resize,m_fixed,h_64,w_64)
论文阅读笔记
Mute杭盖
差距并非一夜产生和拉开,更不是造化弄人,命运只对你不公。差距几乎是在你为自己的人生旋律定调之初就已开始,只是那时你毫无意识,根本想不到这差距有一天会像脱缰的野马,一去不回头。
展开
-
罗浩博士-Bag of Tricks and A Strong Baseline for ReID
摘要这方法简单而有效.别人用复杂的网络和连接多个分支的特征.本文重点在于: 一些有用的training tricks, 他们做的工作就在于collect and evaluate这些有效的training tricks. 他们把这些tricks结合到一起, 在只用全局特征的时候. 模型在Market1501上就达到了超高指标.1. Introductioneffective training tricks or refinementswith involved in all training原创 2020-11-24 07:42:52 · 1246 阅读 · 0 评论 -
CVPR2020-Camera On-boarding for Person Re-identification using Hypothesis Transfer Learning---论文阅读笔记
Abstract他们这个是在to explore the dynamic nature of a camera network, where one tries to adapt the existing re-identification models after on-boarding new cameras(什么是on-boarding new cameras) 意思加入新相机后,还是尝试利用之前存在的re-ID模型。因为最近有a few方法that attempt to address this原创 2020-06-19 05:07:10 · 1180 阅读 · 0 评论 -
Tex2Shape:Detailed Full Human Body Geometry From a Single Image
Abstract程序pipeline梳理原创 2020-06-10 20:20:14 · 844 阅读 · 3 评论 -
TexturePose: Supervising Human Mesh Estimation With Texture Consistency
Abstract这个工作针对的是model-based human pose estimation. Pose estimation具体是指什么? 是字面上的意思么? pose是指姿势么? 还是关键点? 最近取得了较大进展的方法是从图像中直接回归parametric human body model的parameters的方法。 因为图像中没有3D shape ground truth, 相关的方法依赖2D annotations 或者sophisticated architecture designs.原创 2020-06-07 21:44:45 · 424 阅读 · 0 评论 -
MichiGAN:Multi-Input-Conditioned Hair Image Generation for Portrait Editing---论文阅读笔记
Abstract尽管最近用GANs的face image generation很成功,conditional hair editing(什么是有条件的头发编辑)仍是个挑战性问题,由于the under-explored complexity of its geometry and appearance. (human body 还有统一的模式,但是不同发型的话,就压根没有统一范式了。什么短发,长发,卷发,大波浪,小波浪). 在本文中,我们提出多输入conditioned Hair Image GAN, 是原创 2020-06-05 06:27:10 · 630 阅读 · 1 评论 -
Re-Identification Supervised Texture Generation---论文阅读笔记
AbstractIn this paper, we propose an end-to-end learning strategy to generate textures of human bodies under the supervision of person re-identification. (更具体的意思是啥,怎么就在ReID监督下生成human body的texture?)we render the synthetic images with textures extracted fr原创 2020-05-29 06:42:39 · 408 阅读 · 0 评论 -
MicrosoftAsia-Densely Semantically Aligned Person Re-Identification
Abstract为啥牵扯UV空间和texture image因为引入3D body surface后, each position on the 3D body surface(而不是在原来的图像) has a semantic identity (identified by a 2D coordinate$(u,v) $ in UV space) and a texture representation.意思每个点都有对应的语义标识和纹理表达。...原创 2020-05-23 18:07:44 · 185 阅读 · 0 评论 -
MicrosoftAsia-Semantics-Aligned Representation Learning for Person Re-identification---论文阅读笔记和工程实现总结
摘要In this paper, we propose a framework that drives the reID network to learn semantics-alignedfeature representation through delicate supervision designs提出了一个框架,一个驱动reID 网络学习语义对齐的feature represention through 精细的监督设计。学到的特征,在本质上就是语义对齐的意思。...原创 2020-05-19 19:28:09 · 1319 阅读 · 0 评论