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

About RescaleT #45

Closed
Julyliying opened this issue Jul 6, 2020 · 2 comments
Closed

About RescaleT #45

Julyliying opened this issue Jul 6, 2020 · 2 comments

Comments

@Julyliying
Copy link

Xuebin Qin, 您好:
感谢您的分享,真是一项很棒的工作。有一点小疑问想向您请教下,请问您train和test的时候为什么直接把图片resize到(320,320)呢?而不是保持图片的长宽比呢?是跟训练数据有关系还是如果保持长宽比的话,RandomCrop会影响图片中显著性物体的完整性?还是其他什么原因呢?我想训练自己的数据,我的应用场景是1920*1080分辨率的视频,并有一些该分辨率下的训练数据,请问您觉得train和test的size应该怎样调整最合适呢?多谢您的指点!

@xuebinqin
Copy link
Owner

xuebinqin commented Jul 7, 2020 via email

@Julyliying
Copy link
Author

你好,在train的时候把图片resize 到320x320, 一方面是因为batch size在大于一的情况下需要batch内的图像保持统一的spatial size, 另外一方面resize到320x320,有助于network获取更大的感受野(在filter size和pooling operation 一定的情况下),输入图像resize的越小,相对而言感受野越大。在test的时候resize是为了保证其与训练数据在同一scale上。 Randomcrop是为了使得network 对于水平的translation更加的不敏感,从而一定程度上防止过拟合。你可以吧图像resize到320x320,但是分辨率低会对精度有影响,也可以尝试直接用原始分辨率,带来的问题可能是GPU memory不够,同时receptive field相对就变小了,也会影响精度,具体需要尝试,看哪中方式精度高就用哪个。

On Jul 5, 2020, at 9:25 PM, Julyliying @.**> wrote: Xuebin Qin, 您好: 感谢您的分享,真是一项很棒的工作。有一点小疑问想向您请教下,请问您train和test的时候为什么直接把图片resize到(320,320)呢?而不是保持图片的长宽比呢?是跟训练数据有关系还是如果保持长宽比的话,RandomCrop会影响图片中显著性物体的完整性?还是其他什么原因呢?我想训练自己的数据,我的应用场景是19201080分辨率的视频,并有一些该分辨率下的训练数据,请问您觉得train和test的size应该怎样调整最合适呢?多谢您的指点! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#45>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORIMN2WY2ABD47SX2DLR2E73JANCNFSM4ORGI3WQ.

好的。多谢您的解释与回复!!

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