Comments (1)
I found num_samples_cls determines how many samples of one class would present in a batch.
I tried with num_classes=20, batch_size=10 and num_samples_cls=4.
Following is the printed sampled label_id of 5 batches:
batch 1: [15, 15, 15, 15, 7, 7, 7, 7, 9, 9]
batch 2: [9, 9, 0, 0, 0, 0, 4, 4, 4, 4]
batch 3:[17, 17, 17, 17, 1, 1, 1, 1, 6, 6]
batch 4:[ 6, 6, 5, 5, 5, 5, 11, 11, 11, 11]
batch 5:[16, 16, 16, 16, 3, 3, 3, 3, 18, 18]
Don't know why the author chose num_samples_cls to 4. But I guess when batch_size = num_classes * num_samples_cls, this meets the spirit of class-balanced sampling.
from classifier-balancing.
Related Issues (20)
- Problem about ImageNet2014 HOT 1
- question about accuracy HOT 2
- Regarding the Places_LT
- How to use my own data (with different categories and num_classes) to train the model? I found that when I trained directly with the weights you provided, the dimensions did not match HOT 1
- Suit for detection?
- how many gpu do you use?
- Question about weight norm HOT 1
- tau-norm code problems
- How do you implement class-balanced sampling in terms of 'few' class HOT 1
- ImageNet2014 dataset does not correspond to the ImageNet_train.txt
- instance-balanced Training config file for Places365 HOT 5
- tau norm doesn't require --model_dir option?
- Balfeat is not found?
- Shouldn't the Stage1 weights be true in config setting? HOT 1
- More details regarding Places-LT HOT 2
- Learning tau on validation set ? HOT 2
- Resnet10 HOT 3
- How to determine tau? HOT 1
- Cannot get the same acc? HOT 8
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 classifier-balancing.