ICMI'09 Video

In addition to the experience reported at "ICMI-MLMI 2009 at MIT" and "Discussion about Context during the UCVP Workshop", I'm publishing a video showing the visual experience of the ICMI'09 at MIT Media Lab.



Publishing this video, I wish to motivate myself, my colleagues and thank UCL/TELE Lab for this opportunity. It is supposed to be just one more conference, one more place, one more line in my CV, but it is more than all of that. As a simple person, of course, I add an enormous value to this experience and this is a way I found to share what inspired me there.

CEJUG's New J Faces

If we compare the current moment of CEJUG with its first semester in 2009, we would say that the group came back to its old good shape. We have been very active during the last months, thanks to a restructuring process which started in July and since that time it has produced a positive impact on the local Java community. However, this restructuring wouldn't be possible without the direct collaboration of the community. They took the initiative, got mobilised with the coordinators, and became committed to a serious project for the whole community.

We have been repetitively thankful to this group of volunteers in many opportunities, including during last CEJUG's anniversary. Now it's the time to officially recognise them, not only by the leadership team, but for the whole local Java community, who was directly or indirectly impacted. This team was co-responsible for all events we organised since last August and other events this year that are still under negotiation and planning at the moment. They also looked for sponsorship to make 100% of CEJUG's actions possible, for souvenirs to grant our guests and many other things.

The official recognition made by the leadership team was unanimous and we have promoted the following CEJUG's members:

René Araújo becomes JUG Leader in recognition to the voluntary work that he has been doing for the JUG, not only recently, but in several other opportunities during the last years. René is tireless when supporting CEJUG's initiatives. He will join the current team of Jug Leaders composed of Silveira Neto, Rafael Carneiro and Hildeberto Mendonça and also be committed with a voluntary work in benefit of the group indefinitely.

Paulo Júnior becomes events coordinator, role previously assumed by Rafael Carneiro, in recognition of his initiative to leverage the monthly events after a long interruption. He is already managing events since August and he will continue doing that, calling speakers, defining the calendar, negotiating venues and sponsorships, among other responsibilities.

Gregory Fontenele becomes responsible for the CEJUG Certified program and also for the CEJUG Shop in recognition to his great contribution to the last CEJUG's initiatives. Gregory will negotiate sponsorship, promote and reward CEJUG members with vouchers. For the shop, he will organize the order and delivery of t-shirts.

Thiago Sá becomes responsible for the Java Emissary program (literally translated) thanks to his experience on the SUN Campus Ambassador program and in recognition to his collaboration on the last CEJUG initiatives. Thiago will coordinate a network of entrepreneurs undergraduate students who will promote CEJUG in their respective universities.

Francisco Barroso becomes CEJUG supporter, a recognition lower than what he actually deserves, but he has demonstrated his greatness as a person by refusing a more important distinction offered by the leadership team. Francisco is committed to support all CEJUG's initiatives, helping according to his availability.

Felipe Gaúcho becomes a JUG Adviser in recognition to several years of dedication and great achievements for the CEJUG community. Felipe is also responsible for the Java University Award program, which is in its third edition and it is fully sponsored by local companies and supported by the local government and academy. Felipe is not only part of CEJUG's history but he is also helping to design the next steps of the group.

All these activities are a big challenge to these new volunteers. That's why we have to support and remember them for what they did, not for what they didn't do. Even with ups and downs the CEJUG community always comes back stronger and stronger, and you, even without realizing it, are a better developer because of it.

For a Better Lab

My current research experience has taught me some lessons that I would like to share with my readers. For some reason, I had an incredible wish to write about it today and I'm doing it now after getting some inspiration from some Belgian beers.

The point is, how to create a creative environment where people feel really comfortable and motivated to conceive, develop, disseminate and exploit their ideas? I listed below what I believe that could make a research even better, pleasant and productive.

• Research is made by people, not by machines: at least in the area of technology we all work heavily with computers, but all what computers can do is to follow our own instructions, nothing else. So, a research cannot be good enough if it doesn't consider a human or an environmental perspective. So, if you make a bunch of calculations to prove some concept and you don't consider the impact of that proof on people's lives or the environment where the research is inserted, then your research will be less convincing. I don't have enough space to give examples here, but you can test my theory by commenting this post with your questions.
• Being concentrated means being together: sometimes we want to stay alone to get concentrated and have something done. This is totally valid, but the best way to be concentrated is being forced to be concentrated by the presence of other members of your team, if you have a team, of course. These days, there are so many ways to get distracted, reading news, twittering, following friends' lives on Facebook, that a lonely work could be twice unproductive than a cooperative work. Don't be sure that you are doing something together with other people just because you all are in the same research project. Projects can force people to work together and the benefits of being together is only achievable by wishing it.
• Better than publishing is to be referenced: the effort to publish a paper in a good conference or periodic is substantial. We all know that. But the hardest achievement is to have your work referenced by the scientific community. A reference is a recognition that your work exists and had some impact, positive or negative, on somebody else's work. References are also an important indicator of relevance for journals, magazines, conferences and even universities. In the context of a lab, I would say that the best way to get the first references to your work is disseminating your own research in the lab, let people know details about it, and visualize how your solution can be also a solution for other problems in the lab. Usually, researches in the same lab are closely related, increasing the probability that a work can help other works around. Perhaps, some level of envy can be a barrier at the beginning, but you can manage it by taking the initiative of referencing their research first.
• Knowledge management: a lab without knowledge management is like a brain without memory. It reasons many times about the same thing with minor evolution. In fact, companies, that usually don't dedicate most of their time on innovation, are implementing complex and expensive knowledge management solutions. Why labs, where creativity and innovation are the raw material of production, don't do the same? They should. What researchers produce should be easily found and linked with other works. Other researchers should be able to comment on other people's work, recommend further references, suggest improvements, etc. When people bring proceedings from conferences, the whole lab has the right to freely access these resources, but without a knowledge management approach, these innocent gestures are simply forgotten.
• Knowledge sharing: are you afraid to share your ideas in the lab, afraid that people can steal them? Have you considered that your research conclusions might be wrong and the time you spent proving that the wrong is right is twice bigger than discussing your ideas with others and maybe do a shared work with them? Four or more hands produce more than two. So, when you present your work in an internal seminar, make sure that it is a pleasure not an obligation and you are actually looking for contributors, not being free from tough questions.
• There are real scientific contributions between the edges of two or more projects: projects are usually focused on a certain problem or a category of problems, but there is no doubt that the results of a project can produce considerable contributions to other projects. So, it's a sort of blindness when projects just follow their natural course and do not spread a wave of benefits for all other projects around.
• Write alone and write less: if you are desperate about the progress of your research, be sure that one of the causes for this feeling is the fact that you are working mostly alone on your goals. So, find a way to get contribution by contributing for other people's work.

That's it! Be my guest and comment this post because I'm very curious about your opinion. ;-)

Devoxx is a Very Professional Conference

This week I attended, once again, the Devoxx Conference in Antwerp, Belgium. This time I had the opportunity to give a talk there about a software project that I'm working on. Of course, it was a greeeat experience! I have to say that this one was the most professional infra-structure I ever had to give a talk. Amazing! So, congratulations for the whole Devoxx'09 team including organizers, technicians, assistants and others who made our sessions so comfortable, professional and attractive (yeah, I also had my biggest audience ;-) ).

The subject of the talk is not in the scope of this blog, but you can learn more about it in another blog that I contribute to. Here, I'm going to share only the nicest picture of myself in conferences. :D hehehe. Many thanks for the professional photographer there. All the rest you can check out here.

Discussion about Context during the UCVP Workshop

The 4th and 5th days of the ICMI-MLMI 2009 was dedicated to thematic workshops. On the 4th day I attended the Workshop on Use of Context in Vision Processing (UCVP), which was an opportunity for observing recent works on employing contextual information in problems of Computer Vision. There were very good talks all over the day, but what really interested me was a half-hour discussion about "Context" at the end of the workshop.

UCVP workshop venue

At the second floor of the MIT Media Lab, we were freely discussing about why context is important and how it can improve computer vision. The point about context is that if you don't have one, then either your problem is really big or you can solve many problems with a single solution. So, people are used to define a context in order to get results, which is a primary concern of financed research projects, where most of us are allocated on. The independence of context usually comes from attempts to reuse results from previous context-aware works on new projects focused on new contexts. Gradually, the original research becomes more and more generic until it achieves a balanced level of acceptance.

Let me give you an example of a context-aware (context-dependent) research. The work of Stefanie Tellex and Deb Roy, entitled "Grounding Spatial Prepositions for Video Search", was presented during the conference and it considers a database of videos, recorded with a fish eyes camera, showing the routine of a family in a kitchen during some months. They track people and also record what they say to perform a set of experiments. Of course, the experiments are all related to people moving and talking in the kitchen. If they change to a different environment, like a meeting room, the results would change considerably. They cannot even guarantee that their current implementation will work in a different environment. It could be evaluated in other projects and some adjustments on the direction of generalization could be necessary, but still keeping the compatibility with the previous context.

Demo of the domestic surveillance database

Now, an example of a context-unaware (context-independent) research. Antonio Torralba, Rob Fergus, and William T. Freeman have been working on the LabelMe project, a large database of images, with around 80 million tiny images randomly found on the internet. Anyone visiting the LabelMe website can select objects that they recognize and label it with the name of the object. It helps to improve the performance of object recognition algorithms independent of the context. So, they are not studying a specific kind of context, but creating a general solution for several situations of object recognition.

Website of the LabelMe project

What is still confusing for me is that sometimes people mix (or I mix) context and domain. For me, these are two different things, not totally different but differentiable. To make my point clear, let me describe a situation: Imagine a meeting room in a company where directors talk about the strategy of the company. When people go to a meeting at this room we have a contextual situation going on. All concepts within this situation are part of the domain too. But, what about the subjects people are discussing during the meeting? Is it part of the context or is it exclusively part of the domain?

According to the WordNet database, context is a set of facts or circumstances that surround a situation or event; and domain is the content of a particular field of knowledge. Domain seems to embrace context in the sense that it can describe the whole context. According to these definitions, what people talk at the moment is part of the context, but all subjects that people can possibly talk about their business are part of the domain, which should also be considered on the scientific experiments.

So, going back to context-awareness and context-unawareness, are we talking about context or domain? Is it context-aware or domain-aware? Context-unaware or domain-unaware? I never heard of domain-awareness/unawareness, but a lot of context-aware/unaware. How can we keep using the term "context" and still consider what people know but not necessarily say? Open question. Let's think about it ;-)

ICMI-MLMI 2009 at MIT

I'm visiting Boston, USA, right now to attend the Eleventh International Conference on Multimodal Interfaces and the Sixth Workshop on Machine Learning for Multimodal Interaction - ICMI-MLMI 2009. This is the most important conference on the field of Multimodal Interaction, which is one of my PhD's case studies.

Yesterday I made my presentation entitled "A Fusion Framework for Multimodal Interactive
Applications", which is one of the applications of my PhD research. I never spent time in this blog to explain what is multimodal interaction and neither multimodal fusion, but let me do it shortly now: multimodal interaction is the possibility to interact with systems using more than one modality (vision, auditory, touching, etc.) and each modality has a different or complementary influence on the interaction. For example: an text processing software that supports keyboard and also speech recognition as ways to input text. Multimodal fusion is the possibility to use information from two or more modalities, in order to improve a less precise modality or realise more complex meanings by combining data from those modalities. For example: to get better results on speech recognition, a vision system could be used to perform lips reading, improving the probability that a certain word was said in case of uncertainty.

Well, my presentation was good and their questions were tough but they were satisfied with my answers and some of them even sent emails in private, asking for the presentation. I would say that this was the toughest audience that I've ever had. A conference at MIT normally attracts many big names in the field of multimodal interaction. The environment is really appropriate (and magic) to discuss about research and innovation. This year the conference had a big emphasis on robotics, with exciting demos and impressive results on interaction and learning with humans. We also had the chance to make a tour through the lab, visiting different projects and watching live demonstrations (some pictures below).

One of the dozens of rooms in the MIT Media Lab

Social and emotional interaction with a child robot

The experience of visiting the MIT Media Lab was unique. This is a space where people are totally free to create even if ideas sound totally ridiculous. At the end, they find a way to exploit those ideas. It shows that, sometimes, real problems can also block people's creativity and great ideas also come from nowhere.