search

dcharatan / pixelsplat

[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
http://davidcharatan.com/pixelsplat/
MIT License
830 stars 56 forks source link

Evaluation Metric Command #42

Closed jk4011 closed 6 months ago

jk4011 commented 6 months ago

Thank you for your excellent work! Your method creates amazing novel views from just two images.

I'd like to replicate the evaluation metrics, like PSNR, mentioned in your paper. However, the command in the README only generates test images. Could you provide the command for calculating these metrics in your code?

dcharatan commented 6 months ago

To compute evaluation metrics, you can use src.scripts.compute_metrics:

python3 -m src.scripts.compute_metrics +experiment=re10k +evaluation=re10k
python3 -m src.scripts.compute_metrics +experiment=acid +evaluation=acid

You can also use this script to create side-by-side videos if you do +evaluation=re10k_video instead, for example.