[CVPR 2024 Oral, Best Paper Runner-Up] Code for "pixelSplat: 3D Gaussian Splats from Image Pairs for Scalable Generalizable 3D Reconstruction" by David Charatan, Sizhe Lester Li, Andrea Tagliasacchi, and Vincent Sitzmann
Thank you for your absolutely amazing work! It also inspires me a lot.
I have tried to generate gaussians using the code in the folder /paper. And the encoder would generate over 300,000 3dgs based on 2 views. Is there any hyperparameter that could make the model generate one 3dgs per pixel(256*256 3dgs in total)?
I also try to use your compute_psnr code, the PSNR is over 30,which is better than the results in the paper. Is it because in your evaluation the model only generates one 3dgs per pixel?
Thank you for your absolutely amazing work! It also inspires me a lot.
I have tried to generate gaussians using the code in the folder /paper. And the encoder would generate over 300,000 3dgs based on 2 views. Is there any hyperparameter that could make the model generate one 3dgs per pixel(256*256 3dgs in total)?
I also try to use your compute_psnr code, the PSNR is over 30,which is better than the results in the paper. Is it because in your evaluation the model only generates one 3dgs per pixel?
Thank you again for your work!