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

reproduce results #16

Closed
Fungungun opened this issue Jul 12, 2018 · 4 comments
Closed

reproduce results #16

Fungungun opened this issue Jul 12, 2018 · 4 comments

Comments

@Fungungun
Copy link

Hi

I train and test the Market1501 with the default parameters but only got 66% for CMC1 and 43.3% for mAP.

The precision in training is about 99.99%. Do you have any idea what might be wrong?

Cheers

@huanghoujing
Copy link
Owner

Hi, thanks for your attention! I guess the problem is in the version of python and pytorch. Do you use python 2.7 and pytorch 0.3?

@Fungungun
Copy link
Author

Hi

Thanks for your reply.

Problem fixed now I got similar results as reported in this repo with more epochs of training. But it's still strange that since the precision are both around 99.99% and I don't know why the results are so different. Anyway thanks for your work.

@huanghoujing
Copy link
Owner

huanghoujing commented Jul 18, 2018

Glad to hear your results. The number of possible triplets for the training set is huge. So training longer helps the model to see more triplets, which may be useful, although training precision has reached 99%.
In this code, a total of 300 epochs seems satisfying, which is better than 150 epochs. How many epochs did you train that gives you 66% CMC1?

@Fungungun
Copy link
Author

Can't remember. I think I just used the default 150 epochs. I just realized that I stopped and the whole training process in my first try. So I probably have tested with an checkpoint instead of the final trained model. Then it should be my problem. Thanks for your reply and are you on Wechat? If so please add me 382968762

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