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 #3

Open
ym-Rain opened this issue Apr 17, 2021 · 9 comments
Open

Train #3

ym-Rain opened this issue Apr 17, 2021 · 9 comments

Comments

@ym-Rain
Copy link

ym-Rain commented Apr 17, 2021

Thank you for your open source code! If I want to train on MOT20 and MOT15, how can I generate the detection file? Could you help me?

@daip13
Copy link
Owner

daip13 commented Apr 18, 2021

Thank you for your open source code! If I want to train on MOT20 and MOT15, how can I generate the detection file? Could you help me?

You can generate the detection files via this script: ReID_feature_extraction. Of course, you need to make some minor changes about the inputs. Btw, if you want to fine-tune ReID model with MOT20 train data, then you should change the ReID model in config-file.

@ym-Rain
Copy link
Author

ym-Rain commented Apr 18, 2021

Which model should I use? ReID_feature_extraction.py file line 108 of tracktor_prepr_det.txt doesn't exist in the dataset. What shall I do? Thank you for your answer.

@daip13
Copy link
Owner

daip13 commented Apr 18, 2021

Which model should I use? ReID_feature_extraction.py file line 108 of tracktor_prepr_det.txt doesn't exist in the dataset. What shall I do? Thank you for your answer.

You can use the ReID model without MOT train as a baseline. If you want to fine-tune the ReID model, you can refer to fast-reid.
As for the detection txt file, you can use the det.txt file provided by MOT20 benchmark directly. tracktor_prepr_det.txt is the detection results regressed by tracktor.

@ym-Rain
Copy link
Author

ym-Rain commented Apr 18, 2021

Ok, I got it. Thank you for your reply!

@ym-Rain
Copy link
Author

ym-Rain commented Apr 18, 2021

I need to bother you again, I can't run the program( Reid_ feature_ extraction.py) on MOT17. What's the problem? (the Reid model uses Reid_ model_ with_ MOT17_ traindata.pth )
image

@daip13
Copy link
Owner

daip13 commented Apr 18, 2021

I need to bother you again, I can't run the program( Reid_ feature_ extraction.py) on MOT17. What's the problem? (the Reid model uses Reid_ model_ with_ MOT17_ traindata.pth )
image

This script is used to extract the ReID features of each detection box, hence it needs the video image of each frame. You need to merge the images of each sequence to one video first. I believe you can solve such kind of issues easily by yourself after going through the code first.

@ym-Rain
Copy link
Author

ym-Rain commented Apr 18, 2021

If I want to fine-tune the Reid model, which script of fast-reid should I refer to. Is it convenient for you to give me some advice?

@daip13
Copy link
Owner

daip13 commented Apr 19, 2021

If I want to fine-tune the Reid model, which script of fast-reid should I refer to. Is it convenient for you to give me some advice?

https://github.com/daip13/LPC_MOT/blob/master/fast-reid/tools/train_net.py

@ym-Rain
Copy link
Author

ym-Rain commented Apr 19, 2021

Got it. Thank you very much for your guidance!
When I run Reid_ feature_ extraction.py on MOT17, there is an error in the picture below. Do you know what the problem is?
image

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