How do we use the variant of your current models : Ours -im-l2z and Ours-im-pc2z

Question

How do we use the variant of your current models : Ours -im-l2z and Ours-im-pc2z

prashkmr opened this issue 4 years ago · 10 comments

Hi @ChrisWu1997

Can you please tell me how does one use the 2 variants of your models: Ours -im-l2z and Ours-im-pc2z.
Is there any flag or different models made for it, or does one need to manually change the model for running these variants,

I tried searching for some ready-made configuration that can help run these but could not find anyone.

Thanks,
Prashant

Answer 1 · 2020-08-14T01:38:28.000Z

Thanks for taking interest in our paper. Currently the code doesn't support the 2 variants of our model. If you are not in a hurry, I'll update the code to support the 2 variants within a week and inform you once it's done.

Answer 2 · 2020-08-14T07:08:00.000Z

Hi Rundi, Thanks for your prompt reply.Really appreciate it. Actually we ar implementing you paper as a baseline for our work and we were a little short ot time as the conference deadline is near. Would really request to help us with your code as your paper as soon as possible because your paper is a very excellent baseline for ours and we really don't want to miss including it. Thanks, Prashant

…

On Fri, Aug 14, 2020, 07:08 Rundi Wu ***@***.***> wrote: Thanks for taking interest in our paper. Currently the code doesn't support the 2 variants of our model. If you are not in a hurry, I'll update the code to support the 2 variants within a week and inform you once it's done. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEBUDL4X7UFOUMJOBC7JVC3SASISBANCNFSM4P6TE5FA> .

Answer 3 · 2020-08-14T07:27:44.000Z

I see. In this case, I can send you the code for our 2 variants independently (not integrated into this repo). On which dataset do you need? Or it doesn't matter if you are using your own dataset?

Answer 4 · 2020-08-14T13:43:48.000Z

Yes, Actually I am using my own custom dataset which I have setup in the CSV file and in the 2 folders : ShapenetPointCloud and the other one. You can share it with me here.

…

On Fri, Aug 14, 2020, 12:57 Rundi Wu ***@***.***> wrote: I see. In this case, I can send you the code for our 2 variants independently (not integrated into this repo). On which dataset do you need? Or it doesn't matter if you are using your own dataset? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEBUDL7XJ73B2K2UQ6OGMFDSATRP3ANCNFSM4P6TE5FA> .

Answer 5 · 2020-08-14T13:55:54.000Z

Two code folders attached. Feel free to contact me if there are any questions.

…

2020年8月14日下午9:44，PRASHANT KUMAR ***@***.***> 写道： Yes, Actually I am using my own custom dataset which I have setup in the CSV file and in the 2 folders : ShapenetPointCloud and the other one. You can share it with me here. On Fri, Aug 14, 2020, 12:57 Rundi Wu ***@***.***> wrote: > I see. In this case, I can send you the code for our 2 variants > independently (not integrated into this repo). On which dataset do you > need? Or it doesn't matter if you are using your own dataset? > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#1 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AEBUDL7XJ73B2K2UQ6OGMFDSATRP3ANCNFSM4P6TE5FA> > . > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHVOQ3CWBQ2SVSYNUVQTWELSAU5SHANCNFSM4P6TE5FA>.

Answer 6 · 2020-08-14T15:41:08.000Z

Thanks, Chris for this early resolution.
However, I do not see any code folders attached. Can you confirm whether it is in the mail or in the github repo.

Answer 7 · 2020-08-14T16:05:55.000Z

I just resend the code directly to your gmail, please check:)

Answer 8 · 2020-08-14T19:07:01.000Z

Got it, thanks to you, Chris 👍

Answer 9 · 2020-08-14T22:55:41.000Z

Hi Chris,

In the files sent , there is a file common.py that mentions.
self.ae_modelpath = "/mnt/disk4/wurundi/multimodal-pc-completion/3depn-table-pointnetAE-small-2K/model/ckpt_epoch3000.pth"

Does this models refer to the ae or the have trained using the complete Point Clouds?
From your paper, it appears that you use the VAE trained using completely PC in the 2 variants.

Needed a clarification as to whether the line 👍
self.ae_modelpath = "/mnt/disk4/wurundi/multimodal-pc-completion/3depn-table-pointnetAE-small-2K/model/ckpt_epoch3000.pth"

refers to the AE or the VAE trained previously.

Moreover, the 2 variants do not have any scripts for GAN, from which I guess the values of the command line arguments are the default ones assigned in common.py. Correct if I am wrong.

-Prashant

Answer 10 · 2020-08-15T01:14:12.000Z

It refers to the AE pretrained using the complete point cloud. There’s no pretrained VAE involved in our 2 variants. I’ll double check the paper/supp later.

…

在 2020年8月15日，上午6:55，PRASHANT KUMAR ***@***.***> 写道： Hi Chris, In the files sent , there is a file common.py that mentions. self.ae_modelpath = "/mnt/disk4/wurundi/multimodal-pc-completion/3depn-table-pointnetAE-small-2K/model/ckpt_epoch3000.pth" Does this models refer to the ae or the have trained using the complete Point Clouds? From your paper, it appears that you use the VAE trained using completely PC in the 2 variants. Needed a clarification as to whether the line 👍 self.ae_modelpath = "/mnt/disk4/wurundi/multimodal-pc-completion/3depn-table-pointnetAE-small-2K/model/ckpt_epoch3000.pth" refers to the AE or the VAE trained previously. -Prashant — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.