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

I got so low rank1 #26

Open
jinyl777 opened this issue Nov 3, 2019 · 10 comments
Open

I got so low rank1 #26

jinyl777 opened this issue Nov 3, 2019 · 10 comments

Comments

@jinyl777
Copy link

jinyl777 commented Nov 3, 2019

Firstly, thanks for the work and sorry for my low knowledge.
i down the teach model and train the DGnet 100000 Times,but the source was so low ,has any met it and do you give me some suggest

@layumi
Copy link
Contributor

layumi commented Nov 4, 2019

Hi @jinyl777 ,
Thanks for your attention on our work. How about your results? Do you conduct the traditional re-ID evaluation or transfer learning between the datasets?
Could you show me more details? So I could help you to debug the result.

@jinyl777
Copy link
Author

jinyl777 commented Nov 6, 2019

I am honored that you can reply to me, I download the teach model according to the readme, then python train.py --config configs/latest.yaml then test and envlution but got zhe rank1 is 0.001
and then if iwant to train my own data ,what should i do . if you can teach me i weill be honored

@layumi
Copy link
Contributor

layumi commented Nov 6, 2019

@jinyl777
First of all, would you like to evaluate the trained model? Does the trained model achieve the accurate performance?

@jinyl777
Copy link
Author

jinyl777 commented Nov 6, 2019

yes,frist i want trian in my computer and evaluate it , then trian myself dataset

@layumi
Copy link
Contributor

layumi commented Nov 6, 2019

@jinyl777
I mean you may directly use my trained model first. https://github.com/NVlabs/DG-Net#testing

@jinyl777
Copy link
Author

jinyl777 commented Nov 6, 2019

thank you very match , i use this modle i got the nice sorce . if i want to use it to my detaset ,do I have to train my own teach model

@layumi
Copy link
Contributor

layumi commented Nov 6, 2019

Okay. Could you show me your loss curves or some logs? Does the generated image look good?

@jinyl777
Copy link
Author

jinyl777 commented Nov 6, 2019

Because I am still recreating your results, so my data set has not been tested, I will tell you after the test. thanks very math

@jinyl777
Copy link
Author

Hi, I used your code to train the generated image to work well, but loss is not converging.
Epoch: 12 Iteration: 00083796/00100000 DLoss: 2.9820 Reg: 0.0055
L_total: 54.3989, L_gan: 3.0339, Lx: 0.7561, Lxp: 0.7570, Lrecycle:0.3373, Lf: 0.9595, Ls: 0.2201, Recon-id: 7.1530, id: 20.2080, pid:14.4382, teacher: 6.5358
Elapsed time in update: 2.197724

@layumi
Copy link
Contributor

layumi commented Nov 27, 2019

HI @jinyl777 ,
You may modify the re-id learning rate. Try small learning rate.

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