Award Anouncement: The Best Oral is paper: UGLLI Face Alignment: Estimating Uncertainty with Gaussian Log-Likelihood Loss. The Best Poster is paper: Direct Validation of the Information Bottleneck Principle for Deep Nets. Congratulations!

Poster location: The Posters can be put on boards ID 1-19 outside E5. Here is the map.

Workshop location: E5, E6. Oct 27, Full Day.

Sep 30: List of accepted papers are here!

Oct 4: Program details are available! Thank you all, athors, speakers and reviewers!

Oct 4: Instructions on how to prepare poster/oral presentations are here.


Deep learning has been an useful and primary toolbox to perform various computer vision tasks successfully in the recent years. Various seminal works have been proposed to explain the underlying theory and mechanisms of these successful algorithms, in order to further improve their various properties, such as generalization capacity of models, representation capacity of learned features, convergence and computational complexity of training methods.

In this workshop, we consider statistical approaches employed to improve our understanding of deep learning, and to develop methods to boost their properties, with applications in computer vision, such as object recognition, detection, segmentation, tracking, scene description, visual question answering, robot vision, image enhancement and recovery. The workshop will consist of invited talks, oral talks, poster sessions and a research panel. Our target audience is graduate students, researchers and practitioners who have been working on development of novel statistical deep learning algorithms and/or their application to solve practical problems in computer vision. Accepted papers will present their results in the workshop in oral talks and poster sessions. We will also invite selected papers for submission to a special issue on Statistical Deep Learning for Computer Vision in the International Journal of Computer Vision (IJCV). Extended versions of selected papers will be invited for book chapter publication.

Covered Topics

We solicit original contributions that deploy statistical deep learning methods employed to perform various computer vision tasks including, but not limited to:

  • Statistical Understanding of Deep Learning
  • Statistical Normalization Methods
  • Uncertainty in Deep Learning
  • Information Theory of Deep Learning
  • Probabilistic Deep Learning
  • Stochastic Optimization for Deep Learning
  • Probabilistic Programming for Deep Learning
  • Statistical Meta-learning Algorithms
  • Reinforcement Learning for Vision Systems
  • Causal Deep Learning

Invited Speakers

Xianfeng Gu
Stony Brook University
Yi Ma
University of California, Berkeley
Yingnian Wu
University of California, Los Angeles
Lizhong Zheng
Massachusetts Institute of Technology
Alan L. Yuille
Johns Hopkins University

Call for Papers

We invite submission of the extended abstracts (4 pages long excluding references, with optional appendix) describing work in the domains suggested above or in closely-related areas. Accepted papers will be presented during the oral and poster session at the workshop. Workshop papers will appear in the CVF open access archive. Details are listed in the Call for Paper page.

Key Dates:

Paper submission deadline: July 31,2019
Author Notification: Sep 4,2019
Camera-ready deadline: Sep 25,2019