zhichen902 / sc2-pcr Goto Github PK
View Code? Open in Web Editor NEWSC2-PCR: A Second Order Spatial Compatibility for Efficient and Robust Point Cloud Registration (CVPR 2022)
License: MIT License
SC2-PCR: A Second Order Spatial Compatibility for Efficient and Robust Point Cloud Registration (CVPR 2022)
License: MIT License
非常感谢您的开源代码!!我有几个细节的小问题想请教一下
1)您提供了KITTI数据集的FCGF描述子的下载,想请问下每一个点云对的8000对关键点您是随机采样的吗?
2)您的代码中本部分计算correspondence是指在all-to-all的情况下寻找每个点的一个最近邻点吗,也就是关键点的数量=correspondence的数量?
3)如下,您提供的KITTI配置文件中,"num_node": 8000和"max_points": 8000分别是指什么意思呢?哪个值对应提取的关键点数或者是correspondence数量呢?
{
"CUDA_Devices": "1",
"num_iterations": 20,
"dataset": "KITTI",
"ratio": 0.2,
"k1": 30,
"k2": 20,
"data_path": "data/KITTI/fpfh_test",
"descriptor": "fpfh",
"inlier_threshold": 0.6,
"d_thre": 0.1,
"downsample": 0.3,
"re_thre": 5,
"te_thre": 60,
"num_node": 8000,
"use_mutual": false,
"max_points": 8000,
"nms_radius": 0.6
}
您好,非常感谢您的工作,它十分的优秀。在研究您的代码的过程中我发现SC2矩阵计算的过程中使用了两个不同阈值的 SC矩阵,一个是d_thre另一个是d_thre/2。请问这里有什么说法嘛。
The default value is 0.1. It looks like a threshold, but for what? (scaling?) and what are its units?
正常来讲,这六个指标不应该同时变好或者同时变坏的吗?,我主要关注前三个指标,尤其是RR,想请教下您是如何得到论文中的结果的,需要调整哪些参数
2,另外,对内点率这个概念有些疑问,我注意到KITTI FCGF描述子的内点率是30%+,而KITTI FPFH描述子的内点率只有3%+,FCGF描述子和FPFH描述子对应的点云对难道不是一样的吗,如果一样这两个描述子的内点率为什么会不一样
请问作者,我跑您的代码,for循环跑了30次配准,第一次的时间是1.9s,后续的时间都是0.3s左右,是因为一开始是cuda初始化所导致第一次时间很长?可以提取初始化吗?
我觉得您的论文和SAC-COT在**上有一定的相似性,请问作者您和SAC-COT: Sample Consensus by Sampling Compatibility Triangles in Graphs for 3-D Point Cloud Registration这篇论文有进行过比较么?
How to utilize SC^2 to improve the performance of pointDSC? Did you use the SC2_measure from :
Line 353 in e1994d8
and replace the normed_corr_features with SC^2 in :
Hello author, thank you for sharing your work. What's your parameters for compute FPFH in KITTI? e.g. search radius
你好,我想确认一下 该方法使用的训练数据是原始的3DMatch(30%以上重叠率)数据集吗?
the readme tell that I can configure SC2-PRC as:
conda env create -f environment.yml
conda activate SC2_PCR
however, it's missing.
Could you please upload it?
DHVR has not released the code, how did you measure his result?
original words from your paper:The results of DHVR we tested have some difference with the original results, so we also report the results in their pa�per (DHVR-Origin in Tab. 1).
hello,i try to run you code but i found the TE is 79.43, i use you fpfh data.
I want to use initial kitty dataset, what can i do?
作者您好,我想请问一下您测评中KITTI的真值(trans_gt.npy)是如何获得的呢?我下载了KITTI数据,但是发现它提供的位姿真值的数量好像远多于激光点云的帧数,这个对应关系是怎样的呢?
我在我的机子上测试出来pointDSC和你的SC2-PCR 发现结果和论文有比较大差距,并且我实验了下发现利用PointDSC的描述符去补足 空间一致性 导致了更差的结果,不知道您在做实验的时候有没有发现这一点
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.