Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

预处理-实例分割 #30

Open
pengzhongcui opened this issue Oct 6, 2021 · 4 comments
Open

预处理-实例分割 #30

pengzhongcui opened this issue Oct 6, 2021 · 4 comments

Comments

@pengzhongcui
Copy link

您好!我已经运行过您的demo了,想尝试使用其他的kitti数据,同时也会使用别的实例分割算法。因此我很好奇实例分割得到的rgb图怎么能存成物体是1,2,……,n的呢,根据掩膜的区域,以及区域中的颜色吗?如果能分享以下您的代码就好了。万分感谢!

@halajun
Copy link
Owner

halajun commented Jan 8, 2022

@pengzhongcui. 如果是类似mask r-cnn的semantic instance segmentation方法,一般输出结果都有数字标签的,把这些数字标签存储成自己想要的格式就好了。比如我跑mask r-cnn的时候,就在后面输出数据的地方,加几行命令存成自己需要的标签矩阵(背景0或-1,其他物体按顺序标1,2,...,n)就好了。希望帮助到你。

@heng94
Copy link

heng94 commented Apr 9, 2022

@halajun
您好,感谢您分享您的代码。我刚接触这块,想请教您,把mask 存成您说的这种形式有什么优点吗?另外一点,您有重新训练实例分割网络吗?比如mask rcnn,我发现在kitti数据集上的分割效果不怎么理想,想请教您是如何处理这个问题的。
谢谢~

@theyunfeng
Copy link

@pengzhongcui. 如果是类似mask r-cnn的semantic instance segmentation方法,一般输出结果都有数字标签的,把这些数字标签存储成自己想要的格式就好了。比如我跑mask r-cnn的时候,就在后面输出数据的地方,加几行命令存成自己需要的标签矩阵(背景0或-1,其他物体按顺序标1,2,...,n)就好了。希望帮助到你。

请问用了mask rcnn的话,整套slam系统还能保证实时性吗?据我所知maskrcnn本身实时性不好

@Cai-RS
Copy link

Cai-RS commented Sep 19, 2023

@pengzhongcui. 如果是类似mask r-cnn的semantic instance segmentation方法,一般输出结果都有数字标签的,把这些数字标签存储成自己想要的格式就好了。比如我跑mask r-cnn的时候,就在后面输出数据的地方,加几行命令存成自己需要的标签矩阵(背景0或-1,其他物体按顺序标1,2,...,n)就好了。希望帮助到你。

请问用了mask rcnn的话,整套slam系统还能保证实时性吗?据我所知maskrcnn本身实时性不好

你好,可能你还没仔细看论文吧?论文里面说了,实例分割和光流估计是线下进行的,也就是说先进行分割和光流计算,再把这些结果和对应的rbgd图输入系统进行处理。从论文给出的结论来看,不算分割和光流时都达不到10fps,如果再加上这俩只会更慢。当然整个系统没有并行处理,如果有并行,实时性可能会好一些。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants