ACCV 2016 Workshop on
Interpretation and Visualization of Deep Neural Nets
Taipei, Taiwan
Description
Deep Neural Networks (DNNs) are known to excel in many fields, such as image recognition, object detection, video classification, machine translation, reinforcement learning among many others. While their results are impressive, DNNs often acts as a black box and do not provide detailed information about why they reaches a certain classification decision.
Given their nonlinearity and deeply nested structure, such an understanding yields a non-trivial problem, however, it is essential in many practical problems. For instance, in medical diagnosis incorrect prediction can be costly, thus simple black-box predictions cannot be trusted by default. Instead, the predictions should be made interpretable to a human expert for verification.
The Workshop on Interpretation and Visualization of Deep Neural Nets (WINVIZNN2016) encourages submissions related to interpretations, understanding and visualizations of predictions of deep neural nets. We are open about domains, be it theoretical papers which shed light on the success of deep neural nets or interpretation/visualization applications for deep neural nets on images, text documents and videos but also other domains such as audio signals. We welcome – yet are not restricted to - results about neural networks that explain decisions, that compare neural net models, and interpretation/ understanding results for novel application domains. The goal of this workshop is to foster collaboration and understanding on this emerging topic.
Venue
The workshops will be held at Taipei International Convention Center (TICC), the same as the ACCV 2016 main-conference venue.
Publication
The workshop papers will be published by Springer in the Lecture Notes in Computer Science (LNCS) series.
Registration
Workshop registration will be handled as part of the main conference registration. Details will be announced later on both this page and the main conference page.
Schedule
| 13:30-13:50 | Opening + Introductory Talk: Alexander Binder Interpretation and Visualization of Deep Neural Nets |
| 13:50-14:10 | Research Talk: Zhenbing Zhao, Guozhi Xu, Yincheng Qi Multi-Scale Hierarchy Deep Feature Aggregation for Compact Image Representations |
| 14:10-14:30 | Research Talk: Mingming Li, Shuzhi Sam Ge, Tong Heng Lee Glance and Glimpse Network: a Stochastic Attention Model Driven by Class Saliency |
| 14:30-15:00 | Invited Talk: Shixia Liu (Tsinghua University) Towards Better Analysis of Deep Convolutional Neural Networks |
| 15:00-15:30 | Coffee Break |
| 15:30-15:50 | Research Talk: Christoph Käding, Erik Rodner, Alexander Freytag, Joachim Denzler
Fine-tuning Deep Neural Networks in Continuous Learning Scenarios |
| 15:50-16:10 | Research Talk: Ziqin Wang, Peilin Jiang, Fei Wang Dense Residual Pyramid Networks for Salient Object Detection |
| 16:10-16:30 | Research Talk: Nasim Nematzadeh, David M. W. Powers, Trent Lewis Quantitative Analysis of a Bioplausible Model of Misperception of Slope in the Café Wall Illusion |
| 16:30-16:50 | Research Talk: Iaroslav Melekhov, Juho Kannala, Esa Rahtu Image Patch Matching Using Convolutional Descriptors with Euclidean Distance |
| 16:50 | Closing |
Paper Submission
Submissions are required to stick to the ACCV format. Each paper should be at most 14 pages long (excluding references) — shorter papers are welcome, and be submitted into the conference management system. For accepted papers, at least one author must attend the workshop to present the work. The workshop will have an oral and a poster session.
Submission website: https://cmt3.research.microsoft.com/WINVIZNN2016
Important Dates
| Submission deadline | |
|---|---|
| Extended: 3 September, 2016 | |
| Author notification | |
| 13 September, 2016 | |
| Camera-ready | 25 September, 2016 |
| Workshop | 24 November, 2016 |
Organizers
- Wojciech Samek, Fraunhofer Heinrich Hertz Institute
- Alexander Binder, Singapore University of Technology and Design
References
- Erhan et al. Visualizing Higher-Layer Features of a Deep Network
- Yosinski et al. Understanding Neural Networks Through Deep Visualization
- Simonyan et al. Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps
- Zeiler & Fergus. Visualizing and Understanding Convolutional Networks
- Denil et al. Modelling, Visualising and Summarising Documents with a Single Convolutional Neural Network
- Mahendran & Vedaldi. Understanding Deep Image Representations by Inverting Them
- Dosovitskiy & Brox. Inverting Visual Representations with Convolutional Networks
- Nguyen et al. Multifaceted Feature Visualization: Uncovering the Different Types of Features Learned By Each Neuron in Deep Neural Networks
- Zintgraf et al. A New Method to Visualize Deep Neural Networks
- Mordvintsev et al. Inceptionism: Going Deeper into Neural Networks
- Arras et al. Explaining Predictions of Non-Linear Classifiers in NLP
- Yu et al. Visualizing and comparing convolutional neural networks
- Zhou et al. Object Detectors Emerge in Deep Scene CNNs
- Bruckner ML-o-scope: a diagnostic visualization system for deep machine learning pipelines
- Dai et al. Generative modeling of convolutional neural networks
- Vondrick et al. HOGgles: Visualizing Object Detection Features
- Liu et al. Towards Better Analysis of Deep Convolutional Neural Networks
- Harley An Interactive Node-Link Visualization of Convolutional Neural Networks
- Cohen et al. On the Expressive Power of Deep Learning: A Tensor Analysis
- Bach et al. On Pixel-wise Explanations for Non-Linear Classifier Decisions by Layer-wise Relevance Propagation
- Montavon et al. Explaining NonLinear Classification Decisions with Deep Taylor Decomposition
- Samek et al. Evaluating the Visualization of What a Deep Neural Network has Learned