IEEE International Conference on Image Processing 2015
01 Oct 2015
I attended the International Conference on Image Processing (ICIP 2015), in Québec city, Canada, 27-30 September 2015. Our paper that was presented at ICIP2015 was authored by Jie Xie, Michael Towsey, Jinglan Zhang, Xueyan Dong, Paul Roe. It was entitled: Application of Image Processing Techniques for Frog Call Classification.
I did my best to present our research to others during the poster session. I spent two hours standing in front of our poster and explained our work to other researchers. I now realise that a poster is actually a much better way to present your research and learn from others (as opposed to a lecture presentation) because you have enough time to explain to others or learn from others (the poster presentation time was 2 hours per person, whereas the lecture presentation time was only 20 minutes).
After my presentation, four of us with similar interests went to dinner together, which was like another small workshop. This was my favourite part of attending this conference. Another exciting part of this conference trip is that I happened to catch the same airplane with Professor Clinton Fookes on my way there–Clinton also attended ICIP and he was one of the poster session chairs.
I really enjoyed the impressive plenary talks. I was very happy to meet some old friends again there and made some new friends. I also tried to recruit some future research students as there are many Masters students from China attending ICIP2015. Hopefully a couple of them can come to QUT to study PhD one day.
There is great news for next year: The IEEE SPS is doubling the number of travel grants! Many thanks to Prof. Rabab Ward (President-elect of Signal Processing Society) ! PhD students, do your best to apply for the travel grants–I wish many of you a great success!
Plenary talks The opening plenary talk is presented by Professor Yoshua Bengio, head of the Machine Learning Laboratory, Department of Computer Science and Operations Research, University of Montreal, Canada.
Prof. Bengio is most noted for his work on artificial neural networks, especially in deep learning, along with Yann LeCun, Geoffrey Hinton, Andrew Ng et al. He has a vision of how to build computers that can learn like people do. You can read his latest paper here: Deep Learning. Nature 521, 436-444, Yann LeCun, Yoshua Bengio & Geoffrey Hinton, (28 May 2015) doi:10.1038/nature14539
In the plenary talk, Prof. Bengio demonstrated that deep learning is a break-through in the research history of artificial neural networks, which is a branch of machine learning. Deep learning, when well trained, has made many break throughs in wide applications including the processing of video, images, audio and language. Deep learning is promising as it replaces handcrafted features with efficient feature learning algorithms. It can be unsupervised (with unlabelled dataset), supervised (with labelled training dataset), or semi-supervised. Prof. Bengio demonstrated that prior knowledge can speed up the learning process and improve accuracy. Professor Zhenhua Tan, from Aalborg University, Denmark, questioned the conclusion. Prof. Tan said their empirical experience shown that when the dataset is large enough, the effect of prior knowledge almost diminishes, sometimes more accurate (please note: this is only my summary based on my own interpretation, I have no experience on using deep learning).
The second plenary talk, Advances in Computational Imaging, is presented by Prof. Shree Nayar , from Columbia University. He demonstrated that computational imaging is much better than traditional imaging. He showed many recent examples of cameras with latest imaging functionalities (by using new optics). His approach is to manipulate images before they are recorded and processing recorded images before they are presented. It amazed me that Prof. Nayar works on both the hardware and software side of cameras and using cameras to conduct education for children at less developed countries as well. That is truly SYSTEMATIC research!
Québec city is very beautiful, very charming, and very European. Amazingly it has a fortified wall equipped with many cannons, telling us its military history. 95% of Quebec people speak French. You can taste delicious, genuine French food there which you can order in English (Yea!).