What about ICCV2017? Impressions on this year’s event
at
Q: What were your overall impressions of this year’s ICCV conference?
What struck me most was the amount of people. This year’s conference was the biggest ICCV ever, with 3100 attendees. There was half that number at the last one in 2015. It’s grown so much that there was actually an initiative by the PAMI Technical Committee to help manage the numbers. They proposed two motions which were to make the conference annual, instead of biannual, or to make it two track which means you have to make a choice of which orals you want to attend. The community voted for the latter which I’m personally happy about because it’s a solution that has proven to work pretty well in other conferences, like CVPR that has had double tracks since 1991.
Apart from the size, my other global impressions were that it was a very good conference, with great papers and good opportunities for networking and scientific discussions.
Q: What areas attracted the most attention this year?
A lot of good progress is still being made in standard recognition tasks. A good example of this is the COCO challenge where, since last year, detection, segmentation and keypoint localization also improved by about 10%.
The Mask R-CNN method, awarded best paper at the main conference, has a lot to do with this progress, although it was only shared on arXiv a few months ago. This shows the crazy pace of our field.
3D geometry is getting more attention these years, and a lot of what is happening there leverages recent advances in deep learning, proposing “geometry meets deep learning” types of approaches.
Q: Were there any surprises?
Everyone was hiring. Absolutely everyone was hiring and not just for one or two positions – for lots of positions (us included!) The sheer scale of it and the resources companies are putting into hiring is really quite something. Good Ph.D. students can pretty much take their pick from a bunch of earnest employers.
Q: What did you enjoy most at ICCV2017?
The scientific discussions at the posters. It is by far the best way to hold in-depth technical conversations with people who have authored work you like, and of course the opposite way round. People come to see you to learn more about what you’re doing. It’s unfortunate not to be able to do more of this but, because of the number of people there, you often have to queue to chat so it’s only possible to meet a few people at each session if you want to spend some real time with them.
Q: What was the overall feeling about NAVER LABS Europe there?
As we’re new to NAVER LABS but not new to the field, we spent a lot of time sharing what we’re doing in our new home. NAVER is a household name in Korea but is not so well known elsewhere. A lot of their recent investments in Europe as well as the acquisition of our organisation (XRCE) will make that change but we also have a lot to do as scientists in our communities so, you’ll be seeing us more and more at booths and social events as well as of course publishing in the main conferences. ICCV was a great place to start spreading the word.
Q: Do you have any recommendations you could give to young researchers aiming to publish in ICCV for the first time?
It’s such a competitive field that young researchers should most definitely not be discouraged if their paper is rejected. Secondly, when you do have something to publish, make sure you factor in sufficient time for the writing part. Even if you’re proposing an outstanding method, your paper will only be as good as what your community will get out of it. Put yourself in their shoes and your paper will only be better.
About the author: Diane is a senior scientist in the Computer Vision group at NAVER LABS Europe. As well her oral paper, Diane gave an invited talk at the VSM workshop. For details on all NAVER LABS papers and talks at ICCV2017 see this blog post
Our current job openings in computer vision are here
To make robots autonomous in real-world everyday spaces, they should be able to learn from their interactions within these spaces, how to best execute tasks specified by non-expert users in a safe and reliable way. To do so requires sequential decision-making skills that combine machine learning, adaptive planning and control in uncertain environments as well as solving hard combinatorial optimisation problems. Our research combines expertise in reinforcement learning, computer vision, robotic control, sim2real transfer, large multimodal foundation models and neural combinatorial optimisation to build AI-based architectures and algorithms to improve robot autonomy and robustness when completing everyday complex tasks in constantly changing environments.
VISION
The research we conduct on expressive visual representations is applicable to visual search, object detection, image classification and the automatic extraction of 3D human poses and shapes that can be used for human behavior understanding and prediction, human-robot interaction or even avatar animation. We also extract 3D information from images that can be used for intelligent robot navigation, augmented reality and the 3D reconstruction of objects, buildings or even entire cities.
Our work covers the spectrum from unsupervised to supervised approaches, and from very deep architectures to very compact ones. We’re excited about the promise of big data to bring big performance gains to our algorithms but also passionate about the challenge of working in data-scarce and low-power scenarios.
Furthermore, we believe that a modern computer vision system needs to be able to continuously adapt itself to its environment and to improve itself via lifelong learning. Our driving goal is to use our research to deliver embodied intelligence to our users in robotics, autonomous driving, via phone cameras and any other visual means to reach people wherever they may be.
This web site uses cookies for the site search, to display videos and for aggregate site analytics.
Learn more about these cookies in our privacy notice.
Cookie settings
You may choose which kind of cookies you allow when visiting this website. Click on "Save cookie settings" to apply your choice.
This content is currently blocked. To view the content please either 'Accept social media cookies' or 'Accept all cookies'.
For more information on cookies see our privacy notice.