sam. 16 juillet 2016

Translations: en

Hello everyone,

As some of you may know, the IGARSS 2016 conference that held in Beijing (China) just ended. I attended the whole week and I would like to thank the organizers for this great event. This was my first international conference and it already provided amazing memories. Below are some of my thoughts about this year technical content presented at IGARSS, the venue and the spectular city of Beijing.

== Technical content ==

Considering my current research interests, most of the sessions I attended were focused on machine learning (bonus points if it was deep learning !), image processing and data fusion. There was definitely some deep learning hype in the remote sensing community this year, with applications to semantic mapping (cartography) and land cover classification. I was positively surprised to see that many presenters were up-to-date with the bleeding edge litterature in deep learning and I had several interesting discussions on the topic.

However, I found that some presentations had a too narrow scope, presenting only CNN vs SVM results on one dataset. I'd have liked to see more comparisons and experiments justifying the choice of the CNN architecture, the number of layers, the need for fine-tuning, etc. But I'm sure this will come in time and that this year fueled by a deep learning craze that will eventually evolve into sophisticated new approaches of using neural networks for remote sensing data. On that last point, several presentations tried to tackle hyperspectal images classification using deep learning, which is a quite challenging task considering that hyperspectral data contains more bands but fewer pixels than traditional RGB images. And I was thrilled to see that the results are very promising !

I also attended the Sunday tutorial on Big Data Analytics for Geoscience, which was interesting but a little bit to broad in scope to my taste.

== The CNCC (China National Convention Center) ==

IGARSS 2016 venue was located in the North of Beijing in the former Olympic Village. The buildings and facilities are very recent and easy to get to using the subway. The main drawback was that the rooms were not designed to presentations with so many people. Screens were tiny compared to the size of the rooms and it was challenging to read the slides if you were in the back rows. However, the audio system was pretty good and the volunteers were very helpful.

== Beijing ==

Beijing is China's capital and a huge city compared to the European towns I'm used to. Not speaking Chinese proved itself to be challenging sometimes but is a deal breaker. We managed to enjoy chinese food (and beer !) and I took some time to visit a few of the many amazing locations of the city: the Forbidden City and Tian'anmen place, the Heaven Temple and the Lama Temple. The weather was pretty hot (around 30°C) and we even had two sunny days, which was a beautiful surprise as I expected to only see the renowned "pollution grey" of Beijing. The city is incredibly diverse and there are lots of wonderful places to see. This was a really enjoyable location for a conference and I'm glad that I had this opportunity.

Category: conference

Translations: en