mlcc-dclr's Introduction

Multi-Label Confidence Calibration

Official Implementation of our paper:

Tianshui Chen, Weihang Wang, Tao Pu, Jinghui Qin, Zhijing Yang, Jie Liu, Liang Lin.
DCLR: Dynamic Correlation Learning and Regularization for Multi-Label Confidence Calibration.
IEEE Transactions on Image Processing, 2024.

Requirements

We run the code with torch version: 1.11.0, python version: 3.9.7

The detailed requirements is in requirements.txt. Please use the

pip install -r requirements.txt

pip install command to install it.

Datasets

Our method works for COCO and VG datasets, please modify your dataset path under config/dataset. If you want to use your own dataset, you can configure it by copying the dataset writeup under the config/dataset. Afterwards, specify your dataset on the command line using the -dataset parameter. See the config and scripts folders for details on how to configure it.

Usage

# SSGRL module on COCO dataset:
bash scripts/SSGRL_COCO_Calibration.sh

# MLGCN module on COCO dataset: 
bash scripts/MLGCN_COCO_Calibration.sh

# CTran module on COCO dataset:
bash scriptsMLGCN_COCO_Calibration.sh

Citation

@article{chen2024dynamic,
  title={Dynamic Correlation Learning and Regularization for Multi-Label Confidence Calibration},
  author={Chen, Tianshui and Wang, Weihang and Pu, Tao and Qin, Jinghui and Yang, Zhijing and Liu, Jie and Lin, Liang},
  journal={IEEE Transactions on Image Processing},
  year={2024}
}