How to improve the performance further?

[CCOSerika]

Hi, author. I run the code on my own real360 scene and result is like this: To improve the performance, I modify the patchsize from 17 to 33 as the paper says. But the result is out of the expectation. Furthermore, I change the patchsize to 9 and the result is like this. I have no idea what went wrong. Looking forward to ur early reply. Thanx in advance.

[cyz2727327]

Hi, looks like you at least get the real360 scene working, I have been struggling to even start stage1 for it. May I ask your system config and package version please?

[CCOSerika]

Hi, looks like you at least get the real360 scene working, I have been struggling to even start stage1 for it. May I ask your system config and package version please?

Be glad to. Ubuntu 18.04 GPU RTX3090 cuda 11.1 python==3.6.13 flax==0.3.5 jax==0.2.17 jaxlib==0.1.69+cuda111

[cyz2727327]

Hi, looks like you at least get the real360 scene working, I have been struggling to even start stage1 for it. May I ask your system config and package version please?

Be glad to. Ubuntu 18.04 GPU RTX3090 cuda 11.1 python==3.6.13 flax==0.3.5 jax==0.2.17 jaxlib==0.1.69+cuda111

this is awesome, i will try these parameters today, thank you SO MUCH !

BTW, did you change anything in the code besides commenting out the 3 lines requiring 8 GPUs? I have a RTX8000 but it says I need more RAMs

[CCOSerika]

Commenting it out is okey. Besides, some code in stage3 needs to be modified, maybe so.

[cyz2727327]

Commenting it out is okey. Besides, some code in stage3 needs to be modified, maybe so.

Could you share your changed code plz? Thank you so much !

[CCOSerika]

I modified the code for solving the running error and meeting my special requirements. If the same issue occurs, I would share the solution. Or else nothing is to be changed if the code runs smoothly.

[murcje]

Sorry for hijacking this thread, noob here. Got things running with the provided datasets. But how do you go about creating/using own datasets, like you have done?

Thanks!!

[CCOSerika]

Sorry for hijacking this thread, noob here. Got things running with the provided datasets. But how do you go about creating/using own datasets, like you have done?

Thanks!!

Hi, I use LLFF to create own datasets.

[murcje]

Hi, I use LLFF to create own datasets.

Thank you!

[nooskuyh]

Sorry for hijacking this thread, noob here. Got things running with the provided datasets. But how do you go about creating/using own datasets, like you have done? Thanks!!

Hi, I use LLFF to create own datasets.

Can you give me a little more detail about creating an LLIF dataset?

[CCOSerika]

Hi, I use LLFF to create own datasets.

Can you give me a little more detail about creating an LLIF dataset?

Just download the Docker image following the steps provided by LLFF. Then run

$ sudo nvidia-docker run -it --rm --volume /:/host --workdir /host$PWD tf_colmap
$ python imgs2poses.py scene

to create the dataset yourself.

[nooskuyh]

Hi, I use LLFF to create own datasets.

Can you give me a little more detail about creating an LLIF dataset?

Just download the Docker image following the steps provided by LLFF. Then run

$ sudo nvidia-docker run -it --rm --volume /:/host --workdir /host$PWD tf_colmap
$ python imgs2poses.py scene

to create the dataset yourself.

Thanks!! Could you please send me the stage123 and output file(obj, png) that was executed with 3090? it will be of great help.

[CCOSerika]

Thanks!! Could you please send me the stage123 and output file(obj, png) that was executed with 3090? it will be of great help.

So sorry for the late. Here follows my previous result. toydog_phone.zip

[nooskuyh]

Thank you!It will be very helful! Windows용 메일에서 발송된 메일입니다. 보낸 사람: CCOSerika보낸 날짜: 2023년 2월 15일 수요일 오전 11:43받는 사람: google-research/jax3d참조: Smashh712; Comment제목: Re: [google-research/jax3d] How to improve the performance further? (Issue #124) Thanks!! Could you please send me the stage123 and output file(obj, png) that was executed with 3090? it will be of great help.So sorry for the late. Here follows my previous result.toydog_phone.zip—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>