Data

Tools

Indexes

Applications

Experiments

Open Source Science

https://github.com/3mcloud/adafocal

Code for the paper "AdaFocal: Calibration-aware Adaptive Focal Loss"

https://github.com/3mcloud/adafocal

Science Score: 10.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ○
    CITATION.cff file
  • ○
    codemeta.json file
  • ○
    .zenodo.json file
  • ○
    DOI references
  • ✓
    Academic publication links
    Links to: arxiv.org
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (5.6%) to scientific vocabulary
Last synced: 10 months ago · JSON representation

Repository

Code for the paper "AdaFocal: Calibration-aware Adaptive Focal Loss"

Basic Info
Statistics
  • Stars: 8
  • Watchers: 2
  • Forks: 1
  • Open Issues: 0
  • Releases: 0
Created about 3 years ago · Last pushed almost 3 years ago

https://github.com/3mcloud/adafocal/blob/main/ 

# AdaFocal: Calibration-aware Adaptive Focal Loss (NeurIPS 2022)
Official code for the paper **AdaFocal: Calibration-aware Adaptive Focal Loss** 

**Authors**: Arindam Ghosh, Thomas Schaaf, and Matt Gormley 

**URL**: https://proceedings.neurips.cc/paper_files/paper/2022/hash/0a692a24dbc744fca340b9ba33bc6522-Abstract-Conference.html 

**Arxiv**: https://arxiv.org/abs/2211.11838 


The code provides the bare minimum to reproduce the calibration related results obtained from training a ResNet-50 model on CIFAR-10 dataset using Adafocal loss. 


Most of the starter code for training, evaluation and calculating calibration related metrics is borrowed from https://github.com/torrvision/focal_calibration.

## Training
To train Resnet-50 on CIFAR-10 with default settings for Adafocal, run:
```train
python main.py --dataset cifar10 --model resnet50 --loss adafocal -e 350 --save-path exp/cifar10_resnet50_adafocal
```

## Citation
If the code or the paper has been useful in your research, please add the citation:
```citation
@inproceedings{NEURIPS2022_0a692a24,
 author = {Ghosh, Arindam and Schaaf, Thomas and Gormley, Matthew},
 booktitle = {Advances in Neural Information Processing Systems},
 pages = {1583--1595},
 title = {AdaFocal: Calibration-aware Adaptive Focal Loss},
 url = {https://proceedings.neurips.cc/paper_files/paper/2022/file/0a692a24dbc744fca340b9ba33bc6522-Paper-Conference.pdf},
 volume = {35},
 year = {2022}
}
```

Owner

  • Name: 3M
  • Login: 3mcloud
  • Kind: organization
  • Location: Maplewood, MN

Science. Applied to life.

GitHub Events

Total
  • Watch event: 1
  • Fork event: 1
Last Year
  • Watch event: 1
  • Fork event: 1