liufeng2915 / fully_3d_object Goto Github PK

Thank you for sharing the part one code! This is amazing work. Is there any plan on posting the SVR part?

Thanks.

Thanks for your awesome work and code sharing! But I have problem when trying to download your pretrained model from the provided google drive link, it says I need access. When will you open the access? Looking forward to your update!

Looking forward to your second part for SVR.

Hi, thank you for sharing the code of the interesting work. Will you add a license for the code?

Hello, thanks for the great work and extremely well-written paper! It was very easy to follow and a pleasure to read.

Regarding the codebase, I'm wondering how are the AlbedoDecoder and ShapeDecoder trained in Pretain? I referred to /SVR/network.py and it seems like only ImgEncoder is used and trained?

Also, where can we find the implementation of the Finetune function, as defined in /SVR/main.py?

I will also follow-up via email. Thank you and I look forward to referencing your work!

Hi, I'm reading your NeurIPS 2021 paper "Voxel-based 3D Detection and Reconstruction of Multiple Objects from a Single Image". I have some doubts as following:
(1) How to output 3d heatmap and 3d regeression using 2d convs on a 4d tensor?
(2) What's the function of 3d u-net? Is there any ablation study on this?
Looking forward to your reply. Thank you.