Comments (5)
No. I'm trying to understand how the code works and I've tested the brdf_mlp with some random values obtaining very high values (10 e6). Since albedo is between 0 and 1, the differences between scales are strange andthat's the reason I was curious about the brdf_scale but I think is 1.
Is it the expected mlp_brdf to return so high values?
from nerfactor.
I am having a similar problem. I am doing some modifications to the code and now, I see that the brdf_mlp produces very large values ranging from 1e4 to 1e6 or so. This results on the albedo being not optimized as the final RGB value depends mostly on the specular, which has much higher values than the albedo
from nerfactor.
Hi both, that constant for scaling the BRDF is an artificial scale set empirically. The reason is that the unit of BRDFs is often ambiguous and not physically rigorous. For example, when we pretrained our BRDF MLP on the MERL BRDFs, it was unclear what unit the MERL BRDFs used. Plus, the light probe intensities are not metric either, so somewhere along the rendering pipeline, one would need some scaling, whether at the BRDF output or the final rendering. Feel free to reopen this if you need further help.
from nerfactor.
Hello!
I understand your answer but checking the configuration, brdf_scale seems to be 1 so it would be the same to ignore it. Should it be changed for each scene?
Thanks
from nerfactor.
Hi @SirSykon. Shouldn't need to change that for each scene. Usually, we set one brdf_scale
for each BRDF pertaining dataset (e.g., the MERL dataset) because that's what determines the "scale."
from nerfactor.
Related Issues (20)
- How to calculate geometry buffers from MVS geometry ? HOT 1
- Rendering scripts HOT 1
- questions about hdrs. HOT 1
- Question about incompatible shapes(0,3) and (100,3) at II. Joint Optimization in Training, Validation, and Testing HOT 19
- gradient error in Joint Optimization HOT 6
- MLPs wrong skip connection
- Crash at shape pre-training HOT 7
- OOM at II. Joint Optimization in Training, Validation, and Testing HOT 2
- Wrong NeRF and surface
- How long will it take to run the third part in the ./nerfactor HOT 1
- Rendering results are all white after training the vanilla NeRF in step1 HOT 1
- Shape error at II. Joint Optimization HOT 1
- Can we extract mesh from the system by marching cubes? HOT 1
- About create my own dataset
- Relighting Results Background Color HOT 1
- When I train vanilla nerf, there are countless threads.
- It is slow to render my own synthetic data, can we use gpu to render? HOT 1
- Shape pre-trained stage error HOT 2
- How about the light direction and the vier direction?
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from nerfactor.