Comments (4)
Hi, to use data normalization, you have to do normalization during training.
What do you mean "normalization process cannnot output not that great result"? Are you saying the images after normalization are not good or the results after training are not good?
Also are you using the data normalization code provided in this repository?
from ps-fcn.
No I am about to train the data soon.
I made my own normalization code referring the equation you wrote in the paper .
Do you mean the kaiming-normalization in your code ?
from ps-fcn.
No, I don't mean kaiming-normalization. I am not sure whether some of the normalized images look like the ones you showed. But it is possible as the appearance of the normalized images depends on the object reflectance.
If you have implemented the normalization strategy introduced in the paper, I think you can just train the model using the normalized images, and after convergence, the model will learn the mapping from image observation to surface normal. But please remember to multiply the scaling factor if the training image number is different from the testing number.
If your code cannot converge, you may try train the model use my code to see if it works.
from ps-fcn.
Thanks for your advice. I will try it :)
from ps-fcn.
Related Issues (20)
- Cannot unzip the pre-trained model. HOT 2
- Difference between the downloaded model and trained model HOT 3
- How do you get the sphere and bunny dataset? HOT 1
- System RAM not enough, what can I do to decrease memory cost ? HOT 3
- How much is the final loss of the PS-FCN network ? HOT 1
- Question about the feature visualization HOT 2
- about making dataset : HOT 4
- questions about the PS-FCN (B+S+32,16) HOT 4
- Question about max pooling HOT 2
- training dataset HOT 2
- Question obout about light intensity HOT 5
- About DataSets HOT 5
- Hello, I would like to inquire about the synthetic datasets. HOT 5
- How to fix this? TypeError: __init__() missing 1 required positional argument: 'dtype' HOT 2
- about lighting used to make dataset HOT 4
- Normal map to Depth map HOT 2
- About the dataset HOT 1
- about the light direction HOT 4
- coordinate system of light directions HOT 2
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 ps-fcn.