Deep Mixture of Linear Inverse Regressions (CVPR 2017)

Convolutional Neural Networks (ConvNets) have become the state-of-the-art for many classification and regression problems in computer vision. When it comes to regression, approaches such as measuring the Euclidean distance of target and predictions are often employed as output layer. In this paper, we propose the coupling of a Gaussian mixture of linear inverse regressions with a ConvNet, and we describe the methodological foundations and the associated algorithm to jointly train the deep network and the regression function. We test our model on the head-pose estimation problem. From an experimental point of view, our contribution is threefold. First, we show that, at least in the case of head-pose estimation, inverse regression outperforms regression models currently used by state-of-the-art computer vision methods. Second, our method does not require the incorporation of additional data, as is often proposed in the literature, thus it is able to work well on relatively small training datasets. And, finally, our method outperforms state-of-the-art methods in head-pose estimation using a widely used head-pose dataset. To the best of our knowledge, we are the first to incorporate inverse regression into deep learning for computer vision applications.

Deep Mixture of Linear Inverse Regressions Applied to Head-Pose Estimation, Stéphane Lathuilière, Rémi Juge, Pablo Mesejo, Rafael Munoz-Salinas, and Radu Horaud, IEEE Conference on Computer Vision and Pattern Recognition, July 2017

Code: https://github.com/Stephlat/DMLIR

BibTeX:

@inproceedings{lathuiliere:hal-01504847,
  TITLE = {Deep Mixture of Linear Inverse Regressions Applied to Head-Pose Estimation},
  AUTHOR = {Lathuili{\`e}re, St{\'e}phane and Juge, R{\'e}mi and Mesejo, Pablo and Mu{\~n}oz-Salinas, Rafael and Horaud, Radu},
  BOOKTITLE = {{IEEE Conference on Computer Vision and Pattern Recognition}},
  ADDRESS = {Honolulu, Hawaii, United States},
  ORGANIZATION = {{IEEE Computer Society}},
  YEAR = {2017},
  MONTH = Jul,
  PDF = {https://hal.inria.fr/hal-01504847/file/main.pdf},
}