Comments (7)
Hi @amanjain1397 ,
Yes, the data is augmented. Each augmentation is saved as an .npz file, the number of augmentations per .obj is defined by the --num_aug
flag (for example , you can set --num_aug 40
).
I think I listed all the augmentation flags under this line in train_options.py.
--scale_verts
non-uniformly scale the mesh (i.e., in x, y or z)
--slide_verts
percentage of vertices which will be shifted along the mesh surface
--flip_edges
percentage of edges to randomly flip
Also, if you have a way to generate re-meshing of the same object, it would also be a very good data-augmentation. I am looking for scripts to do that, if you have any please feel free to share :)
from meshcnn.
Thanks Rana.
Is the augmentated data used while training the model? Because I see the training mesh size always equal to train .objs files only.
from meshcnn.
Hi @amanjain1397 ,
Sorry I'm not sure I understood?
Let's go through the SHREC classification example from this github. First get the data and run training:
bash ./scripts/shrec/get_data.sh
bash ./scripts/shrec/train.sh
Notice that the flag --num_aug 20
. So look at the data in the alien class (datasets/shrec_16/alien/train
), consider one mesh T133.obj
. Now if you go into alien/train/cache
, you should see 20x augmentations of it (being) saved during training:
T133_000.npz
T133_001.npz
...
T133_019.npz
This happens in mesh_prepare.py
from meshcnn.
Yeah I got this bit Rana.
I just want to confirm that whether the edge features of these augmented meshes are used for training the model or are they just exported as .npz files just for the sake of visualization?
from meshcnn.
Yes of course the augmented data (.npz) are used for training.
Not sure what you mean by visualization of .npz files? But anyway, the only visualization script I provided (util/mesh_viewer.py) takes as input the .obj file. Of course, I only provided how to visualize the test .objs anyway, and by default I do not augment the test dataset.
Does this answer your question?
from meshcnn.
Right, I got it. Thanks.
from meshcnn.
Great, Thanks for letting me know :)
Marking this as closed
from meshcnn.
Related Issues (20)
- IndexError: index out of range
- AssertionError
- What is the relationship between your meshes and original coseg meshes?
- Segmentation of objects with different characteristic
- assert no zero face area
- Does the class "Mesh" have an attribute about face?
- Undestanding .eseg file
- Proper TensorBoard Usage
- How to prepare data for a custom dataset?
- cannot install pytorch=1.2.0 HOT 1
- How to save a segmented (colored) mesh in .obj format?
- IndexError: index out of range
- Using my own obj files for training
- The problem when I run meshcnn classification HOT 1
- Problems with date preprocessing.
- help in figuring out how human_seg dataset was created
- Cannot load pre-trained weights
- Request for further documentation of code options on the meshcnn wiki for the Classification task
- Numpy error? HOT 2
- A small spelling mistake in the paper
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 meshcnn.