08:30 - 08:40 | Welcome and Agenda Presentation |
08:40 - 09:25 | Invited Talk 1: "Recent Advances in Domain Adaptation", Prof. Rama Chellappa Over the last five years or so, methods for adapting features and classifiers to domain shifts have been developed. In this talk, I will present recent methods for domain adaptation using manifold and sparse representations as well as deep features. Remaining issues and challenges will also be discussed. |
09:30 - 9:50 | Invited Talk 2: "Learning Visual Representation from Web Data" Dr. Wen Li |
9:50 - 10:30 | Spotlight Presentation of Workshop Papers and Invited Papers (1-12). 3 minutes each: Paper 1: "Domain adaptive subspace clustering" Mahdi Abavisani Paper 2: "Best Practices for Fine-tuning Visual Classifiers to New Domains" Brian Chu Paper 3: "Deep CORAL: Correlation Alignment for Deep Domain Adaptation" Baochen Sun Paper 4: "Nonlinear Embedding Transform for Unsupervised Domain Adaptation" Hemanth Venkateswara Paper 5: "Unsupervised Domain Adaptation with Regularized Domain Instance Denoising" Gabriela Csurka Poster 6: "Online Heterogeneous Transfer Learning by Weighted Offline and Online Classifiers" Yuguang Yan Paper 7: "Learning the Roots of Visual Domain Shift" Tatiana Tommasi Paper 8: "Heterogeneous Face Recognition with CNNs" Shreyas Saxena Paper 9: "Cross Quality Distillation" Jong-Chyi Su, and Subhransu Maji Paper 10: "Learning Attributes Equals Multi-Source Domain Generalization" Chuang Gan, Tianbo Yang, Boqing Gong Paper 11: "Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation" Muhammad Ghifary, Bastiaan Kleijn, Mengjie Zhang, David Balduzzi, Wen Li Paper 12: "Information Bottleneck Domain Adaptation with Privileged Information for Visual Recognition" Saeid Motiian, Gianfranco Doretto |
10:30 - 11:30 | Poster Session and Coffee Break |
11:30 - 14:00 | Lunch |
14:00 - 14:45 | Invited Talk 3: "Towards Principled Transfer Learning", Prof. Christoph Lampert In contrast to ordinary supervised learning, the theoretical foundations of transfer learning are rather underdeveloped. The few results that do exist generally do not lead to practical algorithms and are therefore largely ignored by practitioners. In the talk I will report about the efforts in my group to develop transfer practical learning algorithms that nevertheless come with theoretical guarantees. As an example, I will discuss recent work on multi-task learning with unlabeled tasks. |
14:50 - 15:35 | Invited Talk 4: "Unifying perspectives on knowledge sharing: From atomic to parameterised domains and tasks", Prof. Timothy Hospedales In this talk I will present a flexible framework for thinking about and modelling knowledge sharing across tasks and domains. This framework encompasses various classic and recent methods as special cases and spans both shallow and deep models. In the process, I will revisit some of the basic assumptions we make in transfer learning including: What is a domain or task anyway? What new capabilities and new questions arise if we define these in different ways? I will answer some of these questions, and highlight others as open for future research. |
15:40 - 16:20 | Spotlight Presentation of Workshop Papers and Invited Papers (13-24). 3 minutes each: Paper 13: "VLAD is Not Necessary for CNN" Xiaojun Wu Paper 14: "Training a Mentee network by transferring knowledge from a Mentor network" Elnaz J. Heravi Paper 15: "Lightweight Unsupervised Domain Adaptation by Convolutional Filter Reconstruction" Rahaf Aljundi Paper 16: "Deep Attributes for One-Shot Face Recognition" Aishwarya Jadhav Paper 17: "Hard Negative Mining for Metric Learning Based Zero-Shot Classification" Maxime Bucher Paper 18: "Transfer Learning for Cell Nuclei Classification in Histopathology Images" Neslihan Bayramoglu Paper 19: "Visual Analogies: A Framework for Defining Aspect Categorization" Penelope Tsatsoulis Paper 20: "Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation" Alexander Kolesnikov Paper 21: "Multi-Task Zero-Shot Action Recognition with Prioritised Data Augmentation" Xun Xu, Timothy Hospedales, Shaogang Gong. Paper 22: "Pixel-Level Domain Transfer" Donggeun Yoo, Namil Kim, Sunggyun Park, Anthony Paek, In So Kweon. Paper 23: "Incremental Classifier and Representation Learning" Sylvestre Rebuffi, Alexander Kolesnikov and Christoph Lampert Paper 24: "Knowledge transfer for scene-specific motion prediction" Lamberto Ballan, Francesco Castaldo, Alexandre Alahi, Francesco Palmieri, Silvio Savarese |
16.20 - 17:20 | Coffee Break + Poster Session B |
17:20 - 17:30 | Best paper announcement and Workshop closing |