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

Train Parameters #43

Open
YangYangGirl opened this issue Oct 29, 2020 · 7 comments
Open

Train Parameters #43

YangYangGirl opened this issue Oct 29, 2020 · 7 comments

Comments

@YangYangGirl
Copy link

YangYangGirl commented Oct 29, 2020

你好,我想请问一下具体是在多少个epoch后做finetune的呢,finetune的学习率和步长具体是多少呢?

我在640 * 640的输入上做训练,使用lr_step为90、120,lr为5e-4,batch_size为8。之后将140的epoch取出,送入第二步做finetune,输入尺寸为800 * 800,使用lr_step为30、60,lr为5e-4,batch_size为8。

训练结果为
image

1.First train with the size of 640×640/514×514

2.Then fine tune with the size of 800×800

3.For the easy and hard part, s = s * np.random. Choice (np.arange(0.3, 1.2, 0.1)). The larger the value, the more small samples will be generated

@YangYangGirl
Copy link
Author

YangYangGirl commented Oct 29, 2020

我想问一下这样的结果正常么,精度上的差距主要是因为缺失第三步骤么?还有这里的border可以带来多少的精度提升呢 :)

s = s * np.random.choice(np.arange(0.3, 1.4, 0.1))

@519571754
Copy link

我想问一下这样的结果正常么,精度上的差距主要是因为缺失第三步骤么?还有这里的border可以带来多少的精度提升呢 :)

s = s * np.random.choice(np.arange(0.3, 1.4, 0.1))

请问,您是否有修改过作者的代码,主要指的是数据增强那块,为什么我按着作者代码的训练调参方式,训练出来的结果精度只有70呢

@YangYangGirl
Copy link
Author

from utils.utils import Data_anchor_sample

去年的实验,我不太记得了。建议把训练的bbox值打出来看一下,此外可以看一下这个函数。

@YangYangGirl
Copy link
Author

设置max_obj=314, k=600能有提升

@YangYangGirl
Copy link
Author

YangYangGirl commented Nov 17, 2021

这是我的调优结果
image

@519571754
Copy link

这是我的调优结果 image

您好,您用v2的精度最高达到多少呢,还有就是K=600是哪个参数,期待您的回复

@519571754
Copy link

这是我的调优结果 image

还有就是是否能够给出您的训练步骤,和关键超参的设置值,比如学习率 batch,是否就是是否按着作者提供的训练步骤进行的训练,谢谢

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