当前位置:网站首页>【NeurIPS 2019】Self-Supervised Deep Learning on Point Clouds by Reconstructing Space
【NeurIPS 2019】Self-Supervised Deep Learning on Point Clouds by Reconstructing Space
2022-04-23 03:47:00 【I'll carry you】
1. 四个问题
1. 解决什么问题
self-supervised learning task for deep learning on raw point cloud data
2. 用了什么方法解决
a neural network is trained to reconstruct point clouds whose parts have been randomly rearranged.While solving this task, representations that capture semantic properties of the point cloud are learned
agnostic of network architecture(与网络架构无关)
3. 效果如何
(没有看懂这一段描述)A linear SVM is trained on the representations learned in an unsupervised manner on the ShapeNet dataset. ?
4. 存在什么问题
这样 随意更改位置预测原来的位置 的 代理任务,真的可以吗?(不过 格子好像比较少,3x3)
2. 论文介绍
While solving this task, representations that capture semantic properties of the point cloud are learned
architecture-agnostic(与网络架构无关的)
In this paper we propose a self-supervised method that learns powerful representations from raw point cloud data.
Our method works by training a neural network to reassemble point clouds whose parts have been randomly displaced.
The key assumption of the proposed method is that learning to reassemble displaced point cloud segments is only possible by learning holistic representations that capture the high-level semantics of the objects in the point cloud.(该方法的关键假设是,只有通过学习捕捉点云中对象高层语义的整体表示,才能学习重新组装置换的点云段。)
具体做法:
5 Discussion
在所有的实验中,我们提出的方法学习到的表示被证明是有效的。这让我们相信,对于点云来说,像[7,21]在图像域中讨论的那样,重建输入任务的简单解决方案并不是一个有意义的问题(就是说,代理任务不能太简单? 任务复杂一点,能学到更好的效果?)
3. 参考资料
4. 收获
Auto-encoder与GAN 有什么区别?
一个新的代理任务:打乱 顺序(3x3),预测原来的位置
5 Discussion
在所有的实验中,我们提出的方法学习到的表示被证明是有效的。这让我们相信,对于点云来说,像[7,21]在图像域中讨论的那样,重建输入任务的简单解决方案并不是一个有意义的问题(就是说,代理任务不能太简单? 任务复杂一点,能学到更好的效果?)
版权声明
本文为[I'll carry you]所创,转载请带上原文链接,感谢
https://blog.csdn.net/weixin_43154149/article/details/124353532
边栏推荐
- MySQL zip installation tutorial
- Basic usage of Google colab (I)
- Basic use of Charles
- Commonly used classes
- Install PaddlePaddle on ARM
- If statement format flow
- Key point detection of human hand based on mediapipe
- Initial experience of talent plan learning camp: communication + adhering to the only way to learn open source collaborative courses
- 将编译安装的mysql加入PATH环境变量
- SQL topic exercise summary
猜你喜欢
AI CC 2019 installation tutorial under win10 (super detailed - small white version)
Now is the best time to empower industrial visual inspection with AI
网络原理 | TCP/IP中的连接管理机制 重要协议与核心机制
[latex] formula group
[AI vision · quick review of robot papers today, issue 28] wed, 1 Dec 2021
Seekbar custom style details
Identificateur, mot - clé, type de données
Cuda11 is installed perfectly in win10 X + pytorch 1.9 (blood flowing into the river) cuda. is_ Available() becomes true!
MySQL zip installation tutorial
SQL learning record
随机推荐
MySQL zip installation tutorial
ROS series (V): common commands in ROS
[string] ranking of country names ----- problem solving notes
Three column layout (fixed width on both sides in the middle and fixed width on both sides in the middle)
ROS series (4): ROS communication mechanism series (4): topic communication practice
SQL topic exercise summary
作为一名码农,女友比自己更能码是一种什么体验?
Section 2 map and structure in Chapter 6
[AI vision · quick review of robot papers today, issue 28] wed, 1 Dec 2021
[AI vision · quick review of today's sound acoustic papers issue 1] Thu, 14 APR 2022
列表、元组、字典和集合的区别
网络原理 | TCP/IP中的连接管理机制 重要协议与核心机制
Source code and update details of new instance segmentation network panet (path aggregation network for instance segmentation)
Writing latex with vscode - the latest tutorial 2022 / 4 / 17
Retrieval question answering system baseline
The super large image labels in remote sensing data set are cut into specified sizes and saved into coco data set - target detection
Basic knowledge of convolutional neural network
Use of rotary selector wheelpicker
ERROR: Could not find a version that satisfies the requirement win32gui
Redis(17) -- Redis缓存相关问题解决