Comments (5)
@xpzbph As an alternative, you could use the weights of 2DPass. I am using that weighting with good results, summarized in the following function.
def weights():
seg_num_per_class = [
0,
55437630,
320797,
541736,
2578735,
3274484,
552662,
184064,
78858,
240942562,
17294618,
170599734,
6369672,
230413074,
101130274,
476491114,
9833174,
129609852,
4506626,
1168181,
]
seg_labelweights = seg_num_per_class / np.sum(seg_num_per_class)
seg_labelweights = np.power(
np.amax(seg_labelweights) / (seg_labelweights + 1e-8), 1 / 3.0
)
seg_labelweights = torch.Tensor(seg_labelweights)
# reduce INF for 'unlabeled' to weight = 0
seg_labelweights[seg_labelweights == float("Inf")] = 0
return seg_labelweights
from cylinder3d.
@L-Reichardt
Thank you for your answer. I have two more questions about the function you provided.
- First, does the array of items inside seg_num_per_class represent the total number of points of each class in the training sequence?
2, it is reasonable to say that the weight of unlabeled should be 0, but according to the function you provided to calculate the weight of unlabeled is not 0?
I hope you can answer my question as soon as possible, thank you very much!
from cylinder3d.
- I presume so. This is from 2DPass not me. Alternatively the KITTI .yaml file also contains the ratios of each point compared to the amount of points in the datatset.
- You can set it to 0 if you want. I am using 0 as the "ignore" class in my loss functions, for this reason it does not matter.
from cylinder3d.
@L-Reichardt
Thank you for your reply, I will continue to make an attempt
from cylinder3d.
@L-Reichardt
Hello, sorry to bother you again, I trained the network with RTX2080ti and used the weight calculation function you provided, but the accuracy decreased after 40 epoch and the loss on the validation set increased, I would like to ask you what is the situation after training?
I look forward to your reply.
from cylinder3d.
Related Issues (20)
- Does it currently support distributed multi card training ? HOT 3
- hello,how to convert onnx model ? how to deploy to embedded ?
- How to load a pretrained model?
- Fix: data_path vor train and val dataloaders are the same
- TypeError: No matching definition for argument type(s) array(uint8, 3d, C), array(int32, 2d, C) HOT 6
- How to output a shape of (N,1) ?
- RuntimeError: radix_sort: failed on 1st step: cudaErrorUnsupportedPtxVersion: the provided PTX was compiled with an unsupported toolchain. HOT 2
- 请问一下kitti的背景类为什么不需要训练
- The performance of the baseline model was 10 percentage lower than Cylinder
- The size of cubic voxel in the paper
- About multi-scan task on SemanticKITTI
- Error when trying to re-train
- Details request HOT 2
- About DDP HOT 1
- How to submit nuscenes results on Test set
- Generalization Experiment-Related Questions
- C++ version?
- about pretrained model
- How to see the demo effect? HOT 1
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 cylinder3d.