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

Questions regarding the training and validation of YOLOv10 #148

Closed
wangqiqi opened this issue May 31, 2024 · 7 comments
Closed

Questions regarding the training and validation of YOLOv10 #148

wangqiqi opened this issue May 31, 2024 · 7 comments

Comments

@wangqiqi
Copy link

  1. 同样的数据集(训练,验证,测试)用YOLOv10训练和 YOLOv8训练,最后的Best 显示,YOLOv8比 YOLOv10要好???
  2. 所有参数都是用默认的;
  3. 都是用的官网 nano 版本作为 pretrained 。
@wangqiqi
Copy link
Author

数据集是私有定制的数据集,不是公开的数据集;
数据集中目标的尺寸变化范围比较大。

@jameslahm
Copy link
Collaborator

您好,感谢您的关注!您好,我们想确认一下您的训练环境是一致的吗?是都基于这个代码库训练YOLOv8和YOLOv10的吗?另一方面,您方便看下这个#111 (comment) 是否有帮助吗?谢谢!~

@wangqiqi
Copy link
Author

训练环境是一致的,YOLOv10的环境是从 YOLOv8 拷贝迁移过来的。环境和数据都是一样的,只有模型不一样。哦,对了,YOLOv8的版本是 8.2.22(pip 升级后的)
YOLOv10的版本是 8.1.34(官网Clone YOLOv10)
所有的超参数都是用默认的,没有改变。预训练模型也都是用官网提供的。

@nobody-cheng
Copy link

或许应该都吧val结果保存下来**--save-json True **, 然后都用这个结果使用cocoapi进行验证比对

@wangqiqi
Copy link
Author

或许应该都吧val结果保存下来**--save-json True **, 然后都用这个结果使用cocoapi进行验证比对

#100 (comment)
#111 (comment)

有可能是定制数据集里小目标比较多的原因吧???

@jameslahm
Copy link
Collaborator

或许应该都吧val结果保存下来**--save-json True **, 然后都用这个结果使用cocoapi进行验证比对

#100 (comment)

#111 (comment)

有可能是定制数据集里小目标比较多的原因吧???

您好,关于小目标请您看一下#136,您的测试评估方式跟阈值有关吗?

@jameslahm jameslahm changed the title 同样的数据集(训练,验证,测试)用YOLOv10训练和 YOLOv8训练,最后的Best 显示,YOLOv8比 YOLOv10要好?是不是哪个地方设置的不对? Questions regarding the training of YOLOv10 Jun 3, 2024
@jameslahm jameslahm changed the title Questions regarding the training of YOLOv10 Questions regarding the training and validation of YOLOv10 Jun 3, 2024
@jameslahm
Copy link
Collaborator

Please feel free to reopen this issue if you have further questions.

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

3 participants