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.

Sign up for GitHub

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

Jump to bottom

Problem about training schedule #5

Closed
GuitarZH opened this issue Mar 7, 2021 · 4 comments
Closed

Problem about training schedule #5

GuitarZH opened this issue Mar 7, 2021 · 4 comments
Labels
clarification Clarification regarding the methodology. will-do Will do in some time.

Comments

@GuitarZH
Copy link

GuitarZH commented Mar 7, 2021 •
edited
Loading

Hello @JosephKJ,
I am confused about the T1~T4 training schedule. There are many training commands in the run.sh, and I have tried to run some experiments. However, I still do not understand the meaning of YAML filenames and which YAML file I should use in each task. Can you release a shell script for reproduction?
Thank you very much!
@ShuoYang-1998
Copy link

I also have the same confusion. If I want to train the T3, should I have to train T1 and T2 models in advance and save the weights? Additionally, should I have to download the provided dataset or just use the originally coco/voc with your provided split txt file?
Thanks!

@GuitarZH
Copy link
Author

GuitarZH commented Mar 8, 2021 •
edited
Loading

Hi @ShuoYang-1998,
I have downloaded the provided dataset "JPEGImages", I think it consists of voc2007~voc2012 and coco dataset. In addition, have you tested the training checkpoint using the visualise_detections.py?

@JosephKJ
Copy link
Owner

JosephKJ commented Mar 9, 2021

If I want to train the T3, should I have to train T1 and T2 models in advance and save the weights?

Yes.

Additionally, should I have to download the provided dataset or just use the originally coco/voc with your provided split txt file?

You can try that. I have basically collapsed every VOC and COCO images and annotations into a single folder. Moving these images across might be time consuming for you, so I have shared it as a single zip file.

In addition, have you tested the training checkpoint using the visualise_detections.py?

No, visualise_detections.py was not used.

However, I still do not understand the meaning of YAML filenames and which YAML file I should use in each task.

Do consider the config files under configs/OWOD. Each t* folder is for each task. Inside use the files as follows: train -> val

Can you release a shell script for reproduction?

Sure. Please excuse me some time to sought out some other compelling things.

@JosephKJ JosephKJ added will-do Will do in some time. clarification Clarification regarding the methodology. labels Mar 9, 2021
@GuitarZH
Copy link
Author

GuitarZH commented Mar 10, 2021 •
edited
Loading

Maybe I misunderstand something. I made it now, thanks.

In addition, have you tested the training checkpoint using the visualise_detections.py?

No, visualise_detections.py was not used.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
clarification Clarification regarding the methodology. will-do Will do in some time.
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@JosephKJ @GuitarZH @ShuoYang-1998