-
Notifications
You must be signed in to change notification settings - Fork 16
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Output looks Poor #16
Comments
Can you post some information? Input images, output images, dataset, training logs, etc. |
Input images are the colmap undistorted output, we had 711 input images. this is the log. It has slang error. == CUDA ==CUDA Version 12.2.0 Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved. This container image and its contents are governed by the NVIDIA Deep Learning Container License. A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience. /ever_training2/ever/splinetracers/slang/optix-intrinsics.slang(509): warning 41010: non-void function does not return in all cases /ever_training2/ever/splinetracers/slang/optix-intrinsics.slang(509): warning 41010: non-void function does not return in all cases /ever_training2/ever/splinetracers/slang/sh-half.slang(21): warning 30081: implicit conversion from 'float' to 'half' is not recommended /ever_training2/ever/splinetracers/slang/sh-half.slang(21): warning 30081: implicit conversion from 'float' to 'half' is not recommended Optimizing /data/output [ITER 7000] Evaluating train: L1 0.042052949965000155 PSNR 23.632291412353517 [07/03 14:00:21] [ITER 7000] Saving Gaussians [07/03 14:00:21] [ITER 30000] Evaluating train: L1 0.022222084552049638 PSNR 28.31905632019043 [07/03 17:31:47] [ITER 30000] Saving Gaussians [07/03 17:31:47] |
28 PSNR doesn't seem bad. What does the result look like? Ignore the slang warning. Also, using 360 images projected to a perspective image will probably cause issues. Just use the 360 images raw and code a special lens to shoot the rays corresponding to the 360 camera. |
The output is looking very cloudy and smoothed out. Maybe i will try with normal video and see how it works and let you know the results... also i will try with the dataset that you have given in github.. |
@half-potato ![]() splatting: ever training ![]() In addition, the PSNR of ever training is much larger than that of gaussian splatting, but the effect is very poor! |
An EVER "splat" is not the same as a gaussian "splat". It cannot be viewed in a splat viewer. It is likely much higher quality when viewed properly. |
@half-potato |
I have implemented EVER in my computer, but my output looks poor than a 3DGS output. What might be the issue ? I have implemented the docker file.. because I was not able to run the normal files. Can you help me with this case..
The text was updated successfully, but these errors were encountered: