A Comprehensive Analysis of Deep Regression

Stéphane Lathuilière 1 Pablo Mesejo 1 Xavier Alameda-Pineda 1 Radu Horaud 1
1 PERCEPTION - Interpretation and Modelling of Images and Videos
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : Deep learning revolutionized data science, and recently its popularity has grown exponentially, as did the amount of papers employing deep networks. Vision tasks, such as human pose estimation, did not escape from this trend. There is a large number of deep models, where small changes in the network architecture, or in the data pre-processing, together with the stochastic nature of the optimization procedures, produce notably different results, making extremely difficult to sift methods that significantly outperform others. This situation motivates the current study, in which we perform a systematic evaluation and statistical analysis of vanilla deep regression, i.e. convolutional neural networks with a linear regression top layer. This is the first comprehensive analysis of deep regression techniques. We perform experiments on four vision problems, and report confidence intervals for the median performance as well as the statistical significance of the results, if any. Surprisingly, the variability due to different data pre-processing procedures generally eclipses the variability due to modifications in the network architecture. Our results reinforce the hypothesis according to which, in general, a general-purpose network (e.g. VGG-16 or ResNet-50) adequately tuned can yield results close to the state-of-the-art without having to resort to more complex and ad-hoc regression models.
Complete list of metadatas

Cited literature [76 references]  Display  Hide  Download

https://hal.inria.fr/hal-01754839
Contributor : Team Perception <>
Submitted on : Monday, April 15, 2019 - 2:49:57 PM
Last modification on : Thursday, April 18, 2019 - 2:10:41 PM

File

main.pdf
Files produced by the author(s)

Identifiers

Collections

Citation

Stéphane Lathuilière, Pablo Mesejo, Xavier Alameda-Pineda, Radu Horaud. A Comprehensive Analysis of Deep Regression. IEEE Transactions on Pattern Analysis and Machine Intelligence, Institute of Electrical and Electronics Engineers, 2019, 41, pp.1-17. ⟨10.1109/TPAMI.2019.2910523⟩. ⟨hal-01754839⟩

Share

Metrics

Record views

475

Files downloads

332