The recent success of deep learning has shown that a deep architecture in conjunction with abundant quantities of labeled training data is the most promising approach for many vision tasks. However, annotating a large-scale dataset for training such deep neural networks is costly and time-consuming, even with the availability of scalable crowdsourcing platforms like Amazon’s Mechanical Turk. As a result, there are relatively few public large-scale datasets (e.g., ImageNet and Places2) from which it is possible to learn generic visual representations from scratch.
Thus, it is unsurprising that there is continued interest in developing novel deep learning systems that trained on low-cost data for image and video recognition tasks. Among different solutions, crawling data from Internet and using the web as a source of supervision for learning deep representations has shown promising performance for a variety of important computer vision applications. However, the datasets and tasks differ in various ways, which makes it difficult to fairly evaluate different solutions, and identify the key issues when learning from web data.
This workshop aims at promoting the advance of learning state-of-the-art visual models directly from the web, and bringing together computer vision researchers in this field. To this end, we release a large scale web image dataset named WebVision or visual understanding by learning from web data. The datasets consists of 16 million of web images crawled from Internet for 5,000 visual concepts. A validation set consists of around 290K images with human annotation will be provided for the convenience of algorithmic development.
Based on this dataset, we also organize the 3rd Challenge on Visual Understanding by Learning from Web Data. The final results will be announced at the workshop, and the winners will be invited to present their approaches at the workshop. An invited paper tack will also be included in the workshop.
News 10.06.2019: Result of WebVision 2019 is released. Congratulations to all participants! Detailed result of each entry for each team will be provided soon.
News 12.04.2019: An FAQs page is online. A few frequently asked questions regardig the restrctions on data usage are explained in details. Drop us an email if you have other questions.
News 03.04.2019: More benchmark models are being released. Check the github page for updates
News 03.03.2019: A benchmark model based on ResNet-50 is released for reference, which achieves 71.49% top5 accuracy on the validation set. Thank Mr. Qin Wang for producing this benchmark model.
News 26.02.2019: The WebVision 2019 challenge will start on March 1st, 2019.
News 03.01.2019: The workshop website is now online.
Start Time | Event |
8:30 | Opening Remarks |
8:40 | Invited Talk: Exploring the Limits of Weakly Supervised Pretraining , Dr. Laurens van der Maaten (Facebook AI Research) |
9:20 | Challenge Overview |
10:00 | Coffee Break |
10:30 | Participant Presentation by Alibaba-Vision ( Alibaba Group) |
10:50 | Participant Presentation by BigVideo (SenseTime) |
11:10 | Poster Session
|
Lunch Break | |
14:00 | Invited Talk: Learning about Fashion from Web Photos, Prof. Kristen Grauman (University of Texas at Austin) |
14:40 | Invited Talk: Deep Learning with Noisy Supervision, Prof. Ivor W.H. Tsang (University of Technology Sydney) |
15:20 | Participant Presentation by huaweicloud (Futurewei, University of Electronic Science and Technlogy China (UESTC), Huawei Cloud, and Xidian Univesity) |
15:40 | Award Session & Closing Remarks |
Challenge Launch Date | March 1, 2019 |
Challenge Submissions Deadline | June 7, 2019 |
Challenge Award Notification | June 10, 2019 |
Paper Submission Deadline | May 15, 2019 |
Paper Notification | May 30, 2019 |
Workshop date (co-located with CVPR'19) | June 16, 2019 |
All deadlines are at 23:59 Pacific Standard Time.