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

真实深度图测试 #3

Open
popper0912 opened this issue Apr 24, 2019 · 3 comments
Open

真实深度图测试 #3

popper0912 opened this issue Apr 24, 2019 · 3 comments

Comments

@popper0912
Copy link

你好 谢谢 你们的创作。 最近研究这篇论文,自己的深度图测试不对 按照你文中的预处理 结果不对,可否给些好建议, 谢谢!!!

@strawberryfg
Copy link
Owner

The preprocessing code was not written by me. AFAK xingyi followed the code of DeepPrior. I took a glance and I think the code is here https://github.com/moberweger/deep-prior/blob/master/src/data/dataset.py (line 98-110). I am not sure if Xingyi can find his preprocessing code.

You may follow the python code of DeepPrior and DeepPrior++, which are standard practice.

@popper0912
Copy link
Author

非常感谢 我是按照 Xingyi 训练数据的代码 https://github.com/xingyizhou/DeepModel/blob/master/training/GetH5DataNYU.py 做的预处理。请允许我做一下简单的描述,给我些建议:
1、我用 Xingyi 的 GetH5DataNYU.py 对 NYU Hand Data 进行了训练;
2、在训练过程中,预处理过程如下:
(1)选取手掌的标记点,这里选取了 ID = 34 ,然后以此点为基准点作为空间的中心点,前后左右上下制作包围盒,最后对应到原图将手部 Crop 出来;
(2)Crop 后的深度图减去 ID 对应的深度值,除以包围盒的一半,归一化到 [-1, 1] 之间;
(3)标签也安装此方法归一化到 [-1, 1]之间,进行训练得到训练模型。

此处,对自己的深度数据进行测试:

我使用某深度相机进行采集测试数据,预处理如下:
(1)使用检测网络将手部检测到,获取手在深度图的位置(矩形框);
(2)然后在深度图中找到手部的质心位置;
(3)以此质心点为基准点将手部 Crop(此过程使用某相机的内参),并使用与训练过程一样的操作将其归一化为 [-1, 1]之间;
(4)最后进行实验测试。

最后得到的结果很不好,基本都是乱的回归点,是否与质心(基准点)选取有关,很期待你的回复,谢谢!!!

@popper0912
Copy link
Author

The preprocessing code was not written by me. AFAK xingyi followed the code of DeepPrior. I took a glance and I think the code is here https://github.com/moberweger/deep-prior/blob/master/src/data/dataset.py (line 98-110). I am not sure if Xingyi can find his preprocessing code.

You may follow the python code of DeepPrior and DeepPrior++, which are standard practice.

Hi, wish your reply. Thank u so much.

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

2 participants