nhshin-mcl / MWR

[CVPR 2022] Moving Window Regression: A Novel Approach to Ordinal Regression

Official Pytorch Implementation of the CVPR 2022 paper, "Moving Window Regression: A Novel Approach to Ordinal Regression."

Paper

A novel ordinal regression algorithm, called moving window regression (MWR), is proposed in this paper. First, we propose the notion of relative rank (ρ-rank), which is a new order representation scheme for input and reference instances. Second, we develop global and local relative regressors (ρ-regressors) to predict ρ-ranks within entire and specific rank ranges, respectively. Third, we refine an initial rank estimate iteratively by selecting two reference instances to form a search window and then estimating the ρ-rank within the window.

The full paper can be found via the link above.

Datasets

• MORPH II
• CLAP2015
• CACD
• UTK

Dependencies

• Python 3
• Pytorch

Preprocessing

We use MTCNN for face detection and face alignment code provided from pyimagesearch for face alignment.

Pretrained Models and Reference Lists

You can download models, IMDB-WIKI pretrained model, and reference lists here.

Train

To train MWR, run the script in train_code folder.

python train_code/train.py

Test

Use the following command for evaluation.

python test_code/op.py --dataset Dataset --regression Regression_type --experiment_setting Experimental_setting --im_path Image_path

Cite

@inproceedings{shin2022moving,
  title={Moving window regression: {a} novel approach to ordinal regression},
  author={Shin, Nyeong-Ho and Lee, Seon-Ho and Kim, Chang-Su},
  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
  year={2022}
}

Other Order Learning Algorithms

• OL (ICLR 2020) paper code
• DRC-ORID (ICLR 2021) paper code
• Chainization (ECCV 2022) paper code
• GOL (NeurIPS 2022) paper code

License

See MIT License